@jentrix/cli 0.5.13 → 0.5.15

package/surface.json

@@ -11,7 +11,6 @@
       },
       "description": "Claim a work lease on a target before starting work. Returns CONFLICT (with the active lease in error.current) when the target already has a blocking/exclusive lease and no approved takeover (policyEvaluationId) is supplied — so you can wait, take over, hand off, or run in parallel BEFORE expensive execution.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "agentJobId": {
             "anyOf": [
@@ -168,7 +167,6 @@
       },
       "description": "Append cases to an EXISTING eval dataset (admin scope + ADMIN role). Each case cuts a NEW immutable version (prior members preserved + the new case) and advances the head, all in one transaction — a consumer pinned to a prior version keeps resolving its original case set. Returns the head version and per-case append results.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cases": {
             "items": {
@@ -295,7 +293,6 @@
       },
       "description": "Link a project to a board, task, repo, doc, decision, metric, or agent. For connected sessions, a REPO link uses the normalized lowercase \"owner/name\" as targetId (a GitHub URL is accepted and normalized) — the link that satisfies the session repo gate. Adding an edge that already exists is a no-op. Needs the workspace ADMIN role.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "label": {
             "anyOf": [
@@ -377,7 +374,6 @@
       },
       "description": "Link two tasks of ANY board kind in the same workspace (cross-board allowed), e.g. a bug that BLOCKS a deal. Kinds: BLOCKS, DEPENDS_ON, RELATES_TO, DUPLICATES. Rejects BLOCKS/DEPENDS_ON edges that would create a cycle — the error names the cycle path by task key. Adding an edge that already exists is a no-op.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "fromTaskId": {
             "description": "Task ID on the FROM side of the edge (any board kind).",
@@ -419,7 +415,6 @@
       },
       "description": "Jentrix MVP: anchor an OPEN session to a work item + accountable owner after the operator confirmed the wizard's choices. The server builds the alignment snapshot (ids + display names), stores it on the session, and returns it — echo it VERBATIM to the operator. Re-aligning replaces the anchor and writes a visible activity row, never silently. The session's project/workspace are immutable — a different project means a new session. Requires expectedUpdatedAt (CAS). taskId null = \"no task — session-level work\".",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "agentEmoji": {
             "anyOf": [
@@ -512,7 +507,6 @@
       },
       "description": "Archive a board (admin scope + board ADMIN) — the inverse of unarchive_boards. Soft delete: the board and its tasks are retained and unarchive_boards reverses it.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "boardId": {
             "description": "Board to archive.",
@@ -537,7 +531,6 @@
       },
       "description": "Archive a contact (soft delete — disappears from pickers but existing deal links are preserved).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "contactId": {
             "description": "ID of the contact to archive.",
@@ -562,7 +555,6 @@
       },
       "description": "Archive a harness run (admin scope + ADMIN role). TERMINAL-ONLY: a non-terminal run is refused with the shared archive preflight naming cancellation as the way through, so a run can never be archived out from under a claimed attempt. Soft delete; the run and its stages are retained.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "harnessId": {
             "description": "Alias for harnessRunId. Accepted interchangeably — pass either one; if both are given they must be equal.",
@@ -592,7 +584,6 @@
       },
       "description": "Archive a model profile (admin scope + ADMIN role). Soft delete; the profile leaves the active list but is retained. Returns the archived profile.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "modelProfileId": {
             "minLength": 1,
@@ -621,7 +612,6 @@
       },
       "description": "Archive a provider connection (admin scope + ADMIN role). REFUSED while an active worker release or an in-flight attempt depends on it (the shared archive preflight — cannot silently break running work); the audit records the dependency report either way. Soft delete. Returns the archived (redacted) connection.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "connectionId": {
             "minLength": 1,
@@ -650,7 +640,6 @@
       },
       "description": "Archive a task (soft delete — it disappears from the board but is recoverable).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "taskId": {
             "description": "ID of the task to archive (soft delete).",
@@ -675,7 +664,6 @@
       },
       "description": "Bind a TRUSTED current provider session/thread id (from provider lifecycle hooks or app-server context — never model-authored text) to a Jentrix session. With sessionId: late-binds a CLI-started session's provider thread (CAS). Without: attach-creates an ACTIVE session for the current provider thread; repeating with the same tuple converges on the existing session. A competing active binding fails SESSION_ALREADY_BOUND. Requires idempotencyKey.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "connection": {
             "oneOf": [
@@ -816,7 +804,6 @@
       },
       "description": "Attach a first-class artifact (PR/diff/log/screenshot/eval report/run summary/…) linked to a task/project/run/decision/Work Order. EXTERNAL_URL storage takes a url; R2_OBJECT storage requires an uploadGrantId from POST /api/artifacts/upload-grant (REST, same bearer) after PUTting the bytes to its presigned uploadUrl — `jentrix artifact upload <file>` runs that whole flow. Returns the stored artifact record.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "approvalRequestId": {
             "anyOf": [
@@ -1094,7 +1081,6 @@
       },
       "description": "Authorize a communication connector (OAuth) via elicitation. Sends the user a URL-mode elicitation to complete the provider grant IN THEIR BROWSER; the freshly-minted access token is deposited straight into the M4 credential broker by the provider callback and NEVER returns over MCP. On accept, pins the credential onto the source BY REFERENCE and reports { authorized, action, credentialRef } — a reference, never a secret. Requires a client that supports elicitation. Admin scope + ADMIN role.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "sourceId": {
             "description": "The communication source (connector) to authorize.",
@@ -1125,7 +1111,6 @@
       },
       "description": "Authorize the external runner's LOCAL git commit for a validated harness stage — a repo:commit permission check that honors freeze, leases, and policy. AUTHORIZATION ONLY: it mints no credential and approves no gate. The runner performs the local commit itself, then reports the SHA via submit_harness_stage_result; the push/PR is the separate single-use brokered repo:push / repo:pr_create path, and merge is the separate always-human MERGE gate. Only a VALIDATED (or COMMIT_GATE) stage on a non-frozen run can authorize; otherwise CONFLICT/FORBIDDEN. Returns { authorized, harnessRunId, branch }. Admin only.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "branch": {
             "anyOf": [
@@ -1203,7 +1188,6 @@
       },
       "description": "Create up to 50 tasks in one column in a single transaction (all-or-nothing). Returns each new task's id, number, and key in input order.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "columnId": {
             "description": "Target column for ALL created tasks (one column per call).",
@@ -1284,7 +1268,6 @@
       },
       "description": "Move up to 50 tasks to new columns at once (each appended to the end of its destination, in input order). Access is checked per task; valid moves apply in one transaction with per-item results.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "idempotencyKey": {
             "description": "Optional idempotency key: retrying with the same key and arguments within 24h returns the original result instead of creating duplicates. Reusing a key with different arguments fails with CONFLICT.",
@@ -1335,7 +1318,6 @@
       },
       "description": "Update up to 50 tasks at once. Access is checked per task; valid updates apply in one transaction and the per-item result array reports any failures so you can retry just those.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "idempotencyKey": {
             "description": "Optional idempotency key: retrying with the same key and arguments within 24h returns the original result instead of creating duplicates. Reusing a key with different arguments fails with CONFLICT.",
@@ -1421,7 +1403,6 @@
       },
       "description": "Cancel a harness run while PRESERVING its audit history (artifacts, findings, decisions, and activity stay queryable). A terminal state move — not an archive/delete. Records the cancelling actor + reason. Admin only.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "expectedUpdatedAt": {
             "description": "Optimistic concurrency guard: the entity's updatedAt from your last read. If it changed since, the call fails with CONFLICT and the error embeds the current state.",
@@ -1469,7 +1450,6 @@
       },
       "description": "Runner only: claim the current turn of a job, opening an AgentRun and returning its runId plus the routedReleaseId the run must EXECUTE against (the candidate when a RUNNING canary routed this claim into its traffic fraction, otherwise the pointed-to baseline; null for unbound/personal claims). The worker MUST run the routedReleaseId. Atomic — a second claim while one run is live is a CONFLICT. Pass the runId to the role session for its submit, and heartbeat it while it works.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "appRevision": {
             "description": "The worker image's build-stamped app revision (git SHA). Refused fail-closed on mismatch with a stamped control plane.",
@@ -1569,7 +1549,6 @@
       },
       "description": "Clone an immutable agent release into a new CANDIDATE while changing only codeSnapshotRef. The server copies every opaque prompt, tool-profile, skill, policy, runtime, dataset, model, and budget pin without returning or requiring the raw snapshots. The clone must pass a fresh gating eval before activation.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "codeSnapshotRef": {
             "maxLength": 400,
@@ -1605,7 +1584,6 @@
       },
       "description": "Copy a harness template into a NEW template in the same workspace (ADMIN; admin scope) — the way you fork a pipeline to vary it without touching the one that is in use. The copy always lands as a DRAFT even when cloned from a published version: it is new content that nobody has approved yet. Omit versionId to clone the published head (falling back to the draft when unpublished), and omit name for '<original> (copy)'. idempotencyKey is REQUIRED, so a retried CI step does not leave two copies behind.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "harnessTemplateId": {
             "minLength": 1,
@@ -1643,7 +1621,6 @@
       },
       "description": "Diff two versions of the same harness template — what a review of 'what changed between v3 and v4' actually needs. Steps are matched BY KEY, never by index, so a reordered pipeline reads as `moved` with its indices rather than as a wall of unrelated changes; a step reports `added`, `removed`, `modified`, `moved` or `unchanged`, and a modified one lists each changed field with its before/after. `settings` carries changes outside the step list (run defaults, budgets, credential refs). `identical: true` means the two definitions canonicalize to the same bytes.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "fromVersionId": {
             "minLength": 1,
@@ -1677,7 +1654,6 @@
       },
       "description": "Close a session with an honest lifecycle outcome (COMPLETED normal end / INTERRUPTED crash / CANCELLED pre-first-turn abort). The server verifies the trace-part manifest against durable artifacts, stores the deterministic RUN_SUMMARY, and sets captureComplete ONLY when every expected part and the summary are acknowledged — lifecycle and capture integrity stay orthogonal. Requires expectedUpdatedAt (CAS).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "captureError": {
             "anyOf": [
@@ -1996,7 +1972,6 @@
       },
       "description": "Create or configure the workspace's software-company profile (idempotent). Creates or adopts the five boards (Product Intake, Roadmap, Bugs, Support, Company Operations) — seeding Product Intake's exact New/Triage/Decision/Approved/Rejected columns and Company Operations' request columns/labels — and updates non-secret settings (repo/base branch, sources, required checks, review schedule, operator, agent bindings). Rejects a missing, cross-workspace, archived, duplicate, or wrong-kind board binding. Requires the workspace ADMIN role; changing the operator or the required release/security bindings requires OWNER. Pass expectedUpdatedAt for optimistic concurrency. Never stores credentials, tokens, or prompts.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "agentBindings": {
             "additionalProperties": {
@@ -2279,7 +2254,6 @@
       },
       "description": "Convert a board to a different kind (TASKS/CRM/SUPPORT/BUGS/ROADMAP). Requires ADMIN role. Eagerly creates the target kind's sidecar on every non-archived task; columns and other-kind sidecars are preserved. Returns how many sidecars were backfilled.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "boardId": {
             "description": "Board to convert. Discover via list_boards(workspaceId).",
@@ -2316,7 +2290,6 @@
       },
       "description": "Freeze a promoted playbook into a Skill Registry SkillVersion (admin scope + ADMIN role). Snapshots the playbook's VERIFIED executable definition (trigger, context, tools, content-hashed policy pins, eval-gated configuration digest) into an immutable, content-addressed version (origin = playbook) and pins it onto the playbook in one transaction. Content-addressed + idempotent: re-converting an unchanged playbook replays to the SAME version (created: false), never a duplicate. Only a promoted (enabled) playbook can be converted. Returns the skill/version ids, ordinal, and content hash.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "playbookId": {
             "minLength": 1,
@@ -2340,7 +2313,6 @@
       },
       "description": "Create an agent profile (admin scope; ADMIN role for a workspace agent). Omit workspaceId for a personal agent (owner-only). Pins role/provider/runtime/trust/scopes and the governance bindings. Returns the new agent's id.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "allowedProjectIds": {
             "description": "Project routing boundary within a workspace (empty = all).",
@@ -2586,7 +2558,6 @@
       },
       "description": "Cut a new immutable CANDIDATE release for an agent (admin scope + ADMIN role). Pins the full executable configuration — model + provider adapter, worker/sandbox image digests + SDK revision, code snapshot, prompt/tool-profile snapshots (folded by content hash), skillRefs, policy revisions, gating dataset version, and budget defaults — into a canonical configurationDigest. M18.2 worker config (R3/R4): pin the AI worker's registered runtime adapter + provider connection + generation controls + fallback either explicitly (runtimeAdapterKey/providerConnectionId/generationControls/workerFallback) or, preferably, by publishing the agent's SAVED worker draft with publishWorkerDraft:true (the configure→publish flow — reads the draft's prompt/skills/model). Either way the draft is validated server-side: an unsupported control (maxTurns for Codex, a raw temperature) or a connection/model that doesn't belong to the adapter is REFUSED, never hashed opaquely. Creation is ungated and never requires a prior eval (it breaks the create/evaluate circularity); a green gating run later marks it ELIGIBLE and set_active_release activates it. Workspace-governed releases MUST pin the full runtime/workspace set. Returns the created candidate.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "agentProfileId": {
             "minLength": 1,
@@ -2795,7 +2766,6 @@
       },
       "description": "Create a STARTING connected session BEFORE provider launch: validates operator membership, project visibility, and that the project links the checkout repository (refused otherwise — no model is invoked). The project determines and locks the workspace. Requires idempotencyKey (the CLI launch nonce).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "connection": {
             "oneOf": [
@@ -2923,7 +2893,6 @@
       },
       "description": "Create an automation rule (admin scope + ADMIN role). Board triggers: task.created, task.moved_to_column (optional columnId), label.added (optional labelId), due.passed, sla.breached, check.failed (a CI run entered a failing conclusion on a PR-linked task). Harness-lifecycle triggers (act on the run's anchor task): harness.created, harness.stage_ready, harness.stage_validated, harness.stage_committed, harness.finding_created, harness.blocked, harness.completed. Actions include the usual board ops plus start_delivery_workflow (launch a ci_failure_triage loop bound to the failing PR) and start_harness_stage_run (launch the run's next undone stage — harness triggers ONLY, in practice harness.stage_ready; deduped against an active stage job, a no-op while the run is frozen/cancelled/blocked, and never approves a gate). Actions run as the rule's creator with source 'automation' and can NEVER trigger other rules (hard loop guard).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "actions": {
             "items": {
@@ -3368,7 +3337,6 @@
       },
       "description": "Create a board in a workspace with the default columns for its kind (TASKS/CRM/SUPPORT/BUGS/ROADMAP, default TASKS).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "idempotencyKey": {
             "description": "Optional idempotency key: retrying with the same key and arguments within 24h returns the original result instead of creating duplicates. Reusing a key with different arguments fails with CONFLICT.",
@@ -3417,7 +3385,6 @@
       },
       "description": "Add a comment to a task (markdown body, max 20k chars).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "body": {
             "description": "Comment body (markdown, max 20k chars).",
@@ -3464,7 +3431,6 @@
       },
       "description": "Create a workspace-scoped contact (name required; email/phone/company/externalId optional). Contacts are shared by CRM deals and SUPPORT tickets. Returns the new contact ID.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "company": {
             "anyOf": [
@@ -3552,7 +3518,6 @@
       },
       "description": "Create a credential RECORD (admin scope + ADMIN role) — an opaque backend REFERENCE (installationId / token id), NEVER a secret. The secret half of a backend still enters via the sealed UI intake (H3). Returns the record id.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "allowedScopes": {
             "items": {
@@ -3638,7 +3603,6 @@
       },
       "description": "Propose a decision in the Decision Register (status PROPOSED). Any member may propose; decisions are immutable after acceptance (edits append a revision).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "affectedWork": {
             "anyOf": [
@@ -3880,7 +3844,6 @@
       },
       "description": "Create a NEW eval dataset and cut its first immutable versions from `cases` (admin scope + ADMIN role) — one version per case, all in one transaction. Refused with CONFLICT if a dataset with this key already exists (use add_eval_cases). Returns the head version and per-case append results.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cases": {
             "items": {
@@ -4020,7 +3983,6 @@
       },
       "description": "Create a harness run governing the spec→plan→stage→review→commit loop over a task/project/repo scope. Returns the new run's id + slug.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "allowedPaths": {
             "description": "Path globs the run is allowed to touch (empty = unrestricted).",
@@ -4155,7 +4117,6 @@
       },
       "description": "Author a stage directly on a harness run (the hand-authored fallback to the coverage-enforced spec→plan decomposition loop; ADMIN only, admin scope). Appends it after the existing stages (next 1-based ordinal) carrying the SAME full plan a decomposition proposal does — title, acceptance criteria, traced spec criteria, expected branch behavior, suggested tests, risk notes, branch, commit policy. A NON-draft stage gets executable backing: it auto-mints a 1:1 Work Order when workOrderId is omitted, so it can be passed to request_harness_stage_run (the run must be task-anchored and non-terminal, a supplied workOrderId must back the run's anchor task, and the stage must carry at least one acceptance criterion + one suggested test). Pass draftOnly: true for a non-executable placeholder (no Work Order; request_harness_stage_run rejects it as a draft). Returns the new stage's id + number.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "acceptanceCriteria": {
             "default": [],
@@ -4359,7 +4320,6 @@
       },
       "description": "Create a harness template with an initial DRAFT version (ADMIN; admin scope). Omit `definition` for the built-in starter pipeline. A draft may be invalid — that is what drafts are for — but it may never carry secret material: a credential is referenced by alias, and a secret-shaped literal is refused at the door. idempotencyKey is REQUIRED: retrying with the same key and arguments replays the original result instead of creating a second template.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "definition": {
             "additionalProperties": false,
@@ -6964,7 +6924,6 @@
       },
       "description": "Open an incident (escalate a failed run / bad action into structured handling: severity, owner, timeline, affected entities). Returns the created incident.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "affected": {
             "description": "Blast radius: ids of affected tasks/contacts/repos/customers (loose ids, no FK).",
@@ -7102,7 +7061,6 @@
       },
       "description": "Create a model profile for a provider connection (admin scope + ADMIN role): the model identifier + a label, capabilities, an optional free-text pricing note (no numeric rate field), and workspace-default flag. Returns the created (redaction-safe) profile.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "capabilities": {
             "additionalProperties": false,
@@ -7220,7 +7178,6 @@
       },
       "description": "File ONE governed company-operations request on the workspace's Company Operations board (in the 'Requested' column, class-labelled ACCESS / PROCUREMENT / CONTRACT / PEOPLE). Records the requester, accountable owner, due date, requested effect, supporting evidence/source links, and amount/currency when applicable. This is only the intake step: moving the request to 'Approval required' mints exactly one PROPOSED approval Decision (pinning task/policies/credentials/sources/amount); the request can enter Approved only after a human accepts that Decision, and reach Done only with a human-authored result plus an evidence link. Requires a configured profile and workspace membership. Never records any payment, signature, provisioning, or employment change.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "amount": {
             "description": "Amount when applicable (procurement/contract).",
@@ -7312,7 +7269,6 @@
       },
       "description": "Create a project (the container that links boards/tasks/repos/decisions, with owner/constraints/milestones/risks). Admin-gated.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "budget": {
             "anyOf": [
@@ -7490,7 +7446,6 @@
       },
       "description": "Create a provider connection for a registered runtime adapter (ADMIN). Validation is mechanical and load-bearing: a vendor without a registered adapter — Google Gemini today — is refused with an explanation, never silently accepted; a custom OpenAI-compatible endpoint must be a public https:// URL (localhost/private/RFC1918/link-local destinations are refused through the SSRF guard); api_key/custom connections must reference a secret already stored through the protected credential-entry flow (providerSecretId — never a raw secret), while a subscription connection carries no server-side secret. Nothing reads the process environment, so a connection is creatable without editing environment files. Verification always starts 'unverified' — the connection is never presumed healthy. Returns the created (redacted) connection.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "baseUrl": {
             "anyOf": [
@@ -7566,7 +7521,6 @@
       },
       "description": "Dry-run a staged plan (task changes / bulk edits / outbound messages / etc.) BEFORE any external side effect. Returns the simulation with computed diffs + projected side effects. NEVER commits — approve + execute happen on the human approval surface (no execute_simulation tool in v1).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "agentProfileId": {
             "anyOf": [
@@ -7663,7 +7617,6 @@
       },
       "description": "Add a subtask (checklist item) to a task.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "idempotencyKey": {
             "description": "Optional idempotency key: retrying with the same key and arguments within 24h returns the original result instead of creating duplicates. Reusing a key with different arguments fails with CONFLICT.",
@@ -7701,7 +7654,6 @@
       },
       "description": "Create a new task in a column. Returns the new task's ID, per-workspace number, and human-readable key (e.g. STK-42).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "columnId": {
             "description": "Target column. Use list_columns(boardId) to discover.",
@@ -7768,7 +7720,6 @@
       },
       "description": "Create an outbound webhook for a workspace (requires the admin token scope + ADMIN role). Events are HMAC-SHA256-signed POSTs with X-Jentrix-Signature/-Timestamp/-Event/-Delivery-Id headers, duplicated as the pre-rename X-Stacks-* set with identical values; failed deliveries retry with backoff for up to 8 attempts. The signing secret is returned ONCE. URL must be public https.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "events": {
             "description": "Event names to deliver (e.g. task.created, comment.created, harness.stage_committed). Omit or pass [] for ALL events.",
@@ -7912,7 +7863,6 @@
       },
       "description": "Create a Work Order for a task (the versioned brief + Context Pack an agent run launches from). Deterministic skeleton; narrative fields are model-assisted.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "acceptanceCriteria": {
             "description": "Required behavior, tests, and definition of done.",
@@ -8045,7 +7995,6 @@
       },
       "description": "Create a new workspace and become its OWNER (write scope). REFUSED for a workspace-pinned token — a token scoped to one workspace cannot mint another (checked before any write). The name derives a unique slug and a task-key prefix, both write-once. Workspace DELETE stays UI-only.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "idempotencyKey": {
             "description": "REQUIRED idempotency key: creating a workspace demands one, so a retried call replays the original result instead of minting a second workspace. Same key + same arguments within 24h replays the original response; same key + different arguments fails with CONFLICT.",
@@ -8077,7 +8026,6 @@
       },
       "description": "Delete an automation rule (admin scope + ADMIN role). Permanent — the rule and its run history are removed. To pause a rule instead, use set_automation_enabled.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "automationId": {
             "description": "Alias for ruleId. Accepted interchangeably — pass either one; if both are given they must be equal.",
@@ -8106,7 +8054,6 @@
       },
       "description": "Delete a comment permanently. Only the comment's author can delete it.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "commentId": {
             "description": "Comment to delete.",
@@ -8131,7 +8078,6 @@
       },
       "description": "Delete a subtask permanently.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "subtaskId": {
             "description": "Subtask to delete.",
@@ -8156,7 +8102,6 @@
       },
       "description": "Delete an outbound webhook and its delivery history (admin scope).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "webhookId": {
             "description": "Webhook to delete.",
@@ -8181,7 +8126,6 @@
       },
       "description": "Discover model profiles from a connection's server-side /v1/models listing (admin scope + ADMIN role): each listed model becomes an ADVISORY (discovered) profile — discovered metadata is advisory, the runtime capability schema stays authoritative. Subscription connections have no server probe and are refused with the manual-add pointer. Idempotent (re-discovery converges). Returns the discovered count, models, seeded profiles, and a summary.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "connectionId": {
             "minLength": 1,
@@ -8208,7 +8152,6 @@
       },
       "description": "Export a harness template version as a portable `jentrix.harness-template` document (omit versionId for the published head). The document carries credential REFERENCES only — {alias, kind, scopes, description}, with no value field and no workspace-local binding — and the exporter SCANS the whole document with the platform secret detector and FAILS rather than emitting a silently-redacted copy. Feed the result to import_harness_template in another workspace.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "harnessTemplateId": {
             "minLength": 1,
@@ -8236,7 +8179,6 @@
       },
       "description": "Runner only: mark the claimed run failed (e.g. a provider error) — the job goes to BLOCKED for a human. Requires the runId.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "error": {
             "description": "Why the run failed.",
@@ -8270,7 +8212,6 @@
       },
       "description": "Find tasks SEMANTICALLY similar to an existing task (pass taskId) or to free text (pass workspaceId + query) via vector embeddings — catches duplicates and related work that keyword search misses. similarity is cosine in [0,1]; ≥0.8 is a strong duplicate candidate (link with add_task_link kind DUPLICATES). Embeddings refresh every ~5 minutes, so very recent edits may not match yet; falls back to an empty result with a notice when semantic search is not configured.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "limit": {
             "description": "Max results (default 10, max 25).",
@@ -8309,7 +8250,6 @@
       },
       "description": "Freeze a harness run (the run-level human STOP control): immediately rejects new claims, parks submissions, signals frozen on heartbeat so the runner pauses, and fails brokered repo/deploy/cloud credentials closed. Records the freezing actor + reason (redacted) + the prior status to restore on resume. Rejects an already-frozen run (resume first). Admin only.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "expectedUpdatedAt": {
             "description": "Optimistic concurrency guard: the entity's updatedAt from your last read. If it changed since, the call fails with CONFLICT and the error embeds the current state.",
@@ -8355,7 +8295,6 @@
       },
       "description": "Fetch one agent profile: role/provider/runtime/trust, scopes, skills, routing boundaries, cost profile, eval/failure caches, escalation path, status, linked credentials (no secret material), and a computed launch-eligibility breakdown (eligible + per-check reasons). Legacy recentRuns/latestEvals arrays remain empty; use the Agent Job, eval, trace, and release tools for current execution and quality state.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "agentId": {
             "description": "ID of the agent to fetch.",
@@ -8378,7 +8317,6 @@
       },
       "description": "Get one agent-workflow job's full state: phase, whose turn it is, iteration counts, runs, findings (with how-fixed notes), pending approval gates, and the plan versions. This is the poll endpoint — call it to decide whether it is your turn.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "jobId": {
             "minLength": 1,
@@ -8400,7 +8338,6 @@
       },
       "description": "Get one agent release in full: every pin group by content hash (model/runtime/workspace/governance), the field-level diff vs the immediate prior release, the gating EvalRun evidence (status/score/evidence digest/trace), the pointer transition history this release took part in, and its activation waivers. isActive reflects the profile pointer, not the status.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "releaseId": {
             "minLength": 1,
@@ -8427,7 +8364,6 @@
       },
       "description": "Fetch one connected session: lifecycle status, capture integrity (orthogonal to status), repository start/end state, linked summary/transcript artifacts, the stable usage object (provider-reported tokens or null — never estimated), and the append-only task-alignment interval history (alignments[], ordinal-ascending; the session's taskId mirrors the open interval).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "sessionId": {
             "minLength": 1,
@@ -8449,7 +8385,6 @@
       },
       "description": "Get one artifact record plus a short-lived signed downloadUrl (internal R2 objects; provenance-gated) or the external URL — the tool-surface twin of the stacks://artifact/{id} resource, for clients that only call tools. Redacted artifacts resolve to a tombstone with no URL.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "artifactId": {
             "minLength": 1,
@@ -8471,7 +8406,6 @@
       },
       "description": "THE orientation tool: board info + columns + concise task rows in one call. Use this first instead of list_boards → list_columns → list_tasks. Capped at 500 tasks (notice explains how to page the rest).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "boardId": {
             "description": "Board to snapshot. Discover via list_boards(workspaceId).",
@@ -8494,7 +8428,6 @@
       },
       "description": "Fetch one CI run in full, with its individual child check runs (name, status, conclusion, required flag, deep link). Discover ciRunId via list_ci_runs or get_pull_request.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "ciRunId": {
             "description": "CI run id (from list_ci_runs / get_pull_request).",
@@ -8517,7 +8450,6 @@
       },
       "description": "Read the Company Pulse: pure read-only aggregations over the workspace's existing tables (untriaged signals, accepted opportunities, active projects, agent runs, pending approvals, open incidents, support tickets, company-operations approvals/overdue) plus setup/preflight/reconciliation health. This is the SAME measured projection the /[workspaceSlug]/company page renders — every metric carries its definition, time window, and a workspace-relative link to the exact underlying records. Returns { configured: false } when no profile exists. No writes, no new analytics store.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "workspaceId": {
             "minLength": 1,
@@ -8539,7 +8471,6 @@
       },
       "description": "Fetch a Context Pack: the versioned, checksummed source/tool/policy manifest pinned to a Work Order version, plus its token-budget estimate and budget source (router or heuristic).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "contextPackId": {
             "description": "ID of the Context Pack.",
@@ -8562,7 +8493,6 @@
       },
       "description": "Fetch a decision with its full rationale, options, dissent, evidence, supersession lineage, and revision history (accepted decisions are immutable except via revisions).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "decisionId": {
             "description": "ID of the decision to fetch.",
@@ -8585,7 +8515,6 @@
       },
       "description": "Get one eval case in full: fixture, grader kind + config + version, tags, origin, and provenance (incidentId / sourceRunIds / sourceTraceId).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "evalCaseId": {
             "minLength": 1,
@@ -8612,7 +8541,6 @@
       },
       "description": "Get one eval dataset + its immutable, append-only versions (each with its content hash + pinned-case count). A run_eval with no explicit version runs the head; a gate references a specific version so advancing the head never changes what an existing release was gated on.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "datasetId": {
             "minLength": 1,
@@ -8639,7 +8567,6 @@
       },
       "description": "Get one eval run WITH its per-case results (the 'results reads' surface). The run carries its terminal status, aggregate score, gating flag, evidence digest, and trace id; each result row carries its verdict (PASSED/FAILED/ERRORED or the fail-closed EVIDENCE_UNAVAILABLE/EVIDENCE_EXPIRED), grader, and clickable traceId — so a deliberately failing case reads as a legible failed run.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "evalRunId": {
             "minLength": 1,
@@ -8666,7 +8593,6 @@
       },
       "description": "Get one experiment in full: mode, baseline/candidate releases, traffic fraction, guardrail spec (max spend / latency SLO / min quality / incident count), the governance snapshot recorded at start (pointer revision + regression head + required policy revisions — what the sweep compares to detect SUPERSEDED/STALE drift), the gating evidence a canary started on, terminal outcome + reason, and the baseline-vs-candidate comparison report (routed populations + live spend/latency/quality/incident metrics — shadow and canary share it).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "experimentId": {
             "minLength": 1,
@@ -8688,7 +8614,6 @@
       },
       "description": "Fetch a harness run's config, freeze state (frozen flag + freeze block), and the LIVE registry projection (per-stage status, the human gate each stage parks at, branch, commit, evidence count, and done/total) generated from the stage records — not a stored markdown blob. detailed (default) also returns the artifacts, decisions, findings, and delivery-correlation arrays; pass response_format: 'concise' for harness + registry only. This output is intentionally FROZEN and does NOT include pending gates or active/stuck runs — read those through companion calls: list_harness_stages (per-stage gate/branch/commit/evidence), get_agent_job / list_agent_jobs (loop phase + approval gates), and list_control_tower_items (the harness queues, incl. stuck/blocked). Agent-role (planner/reviewer) readiness has no read tool — a run missing roles surfaces as a launch/start failure, and roles are configured via set_harness_agents (whose result/conflict payload reports current assignment) or the harness Settings UI. Human gate approval is UI-first — there is no MCP self-approval tool.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "harnessId": {
             "description": "Alias for harnessRunId. Accepted interchangeably — pass either one; if both are given they must be equal.",
@@ -8724,7 +8649,6 @@
       },
       "description": "Fetch one harness artifact (by its HarnessArtifactLink id from get_harness): kind, version, checksum, supersession lineage, producing agent/model, plus a short-lived signed download URL (internal R2 objects) or external URL. The body is never inlined and the title is redacted — no unredacted secret is ever returned.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "harnessId": {
             "description": "Alias for harnessRunId. Accepted interchangeably — pass either one; if both are given they must be equal.",
@@ -8758,7 +8682,6 @@
       },
       "description": "Fetch one harness template with its full version history: every version's status (DRAFT or PUBLISHED), content hash, and lineage, plus the draft's validation summary INCLUDING the findings that gate publishing. detailed (default) returns each version's definition document — the ordered steps, per-step role assignments, evidence requirements, credential ALIASES (never values), and run-level defaults; pass response_format: 'concise' to omit the documents when you only need publish state. A PUBLISHED version is immutable: editing forks a new draft rather than changing it.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "harnessTemplateId": {
             "minLength": 1,
@@ -8788,7 +8711,6 @@
       },
       "description": "Get one incident in full: timeline, root cause, affected entities, rollback log, policy/eval gaps, linked decisions/policies/credentials/artifacts, and follow-up task ids.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "incidentId": {
             "minLength": 1,
@@ -8810,7 +8732,6 @@
       },
       "description": "Fetch one work lease by id: target, holder, mode, conflict policy, status, expiry, and takeover/handoff lineage.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "leaseId": {
             "minLength": 1,
@@ -8832,7 +8753,6 @@
       },
       "description": "Get one governed memory item in full: kind/scope, lifecycle status + `selectable`, title/body, provenance (source kind + id), the nominating/promoting actors + review note, supersession lineage, and TTL. Dates are ISO strings.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "memoryItemId": {
             "description": "The MemoryItem id to read in full.",
@@ -8860,7 +8780,6 @@
       },
       "description": "Fetch one recorded policy evaluation by id: the action class, final outcome, machine+human reason, the matched policy (if any), and the REDACTED action inputs. Use this to explain why an action was allowed/denied/required approval.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "evaluationId": {
             "minLength": 1,
@@ -8882,7 +8801,6 @@
       },
       "description": "Fetch a project with its owner, links, milestones, risks, and counts. Accepts projectId, or workspaceId + slug.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "projectId": {
             "description": "Project ID. Provide projectId, or workspaceId + slug.",
@@ -8910,7 +8828,6 @@
       },
       "description": "Fetch one delivery PR link in full: PR fields (state, draft, review decision, mergeable, head/base branch), owning task, CI runs with check counts, redacted CI-evidence count, and the GitHub deep link. Discover prLinkId via list_delivery_items.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "prLinkId": {
             "description": "Delivery PR link id (from list_delivery_items).",
@@ -8933,7 +8850,6 @@
       },
       "description": "Get one simulation: the staged plan, its checksum, and projectedEffects (diffs + projected notifications/activities/webhooks/automations/cost/credentials/approvals/lease-conflicts computed BEFORE any external side effect).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "simulationId": {
             "minLength": 1,
@@ -8955,7 +8871,6 @@
       },
       "description": "Get one skill in full: its identity + mutable head pointer and every IMMUTABLE version (content hash, lifecycle status, isHead flag, and the executable snapshot that IS the hashed content). Editing a skill never mutates a version — a new one is cut — so this history is append-only. Dates are ISO strings.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "skillId": {
             "description": "The Skill id to read (with its versions).",
@@ -8983,7 +8898,6 @@
       },
       "description": "Read the workspace's single software-company profile: its non-secret settings (operator, repo/base branch, required check names, weekly-review schedule, enabled/paused state), the resolved five-board + per-harness agent bindings, and the last reconcile result. Returns { configured: false } when no profile exists. Never returns credentials, tokens, or prompts.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "workspaceId": {
             "minLength": 1,
@@ -9005,7 +8919,6 @@
       },
       "description": "Fetch one spec with all its immutable versions, including per-version quality warnings (vague objective / missing acceptance / missing test plan / conflicting constraints) and a 0-100 quality score.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "specId": {
             "minLength": 1,
@@ -9027,7 +8940,6 @@
       },
       "description": "Fetch full details of a task including description, labels, assignees, subtasks, links, and recent activity. Accepts taskId, or workspaceId + number (the numeric part of a task key like STK-42). Pass response_format: 'concise' for scalars + counts only.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "number": {
             "description": "Task number within the workspace (numeric part of the key).",
@@ -9066,7 +8978,6 @@
       },
       "description": "Report THIS token's context (read scope) so an agent fails fast on scope instead of probing with FORBIDDEN calls: effective + raw stored scopes (empty stored = grandfathered full access until rotated), workspace pinning, displayName/emoji identity, expiry, and the per-minute rate-limit window (limit / used / remaining / reset). NO bearer material is ever returned.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {},
         "type": "object"
       },
@@ -9080,7 +8991,6 @@
       },
       "description": "Fetch a Work Order with its draft fields (objective, acceptance criteria, non-goals, constraints, evidence, escalation), status, current immutable version, and the latest launch (pinned version + context pack ids).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "workOrderId": {
             "description": "ID of the Work Order.",
@@ -9103,7 +9013,6 @@
       },
       "description": "Fetch ONE immutable Work Order version's full frozen body (objective, acceptance criteria, non-goals, constraints, evidence requirements, escalation path) plus its Context Pack reference and mint-time checksum. Resolve by workOrderVersionId, or workOrderId + versionNumber.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "versionNumber": {
             "maximum": 9007199254740991,
@@ -9134,7 +9043,6 @@
       },
       "description": "Runner only: extend a claimed run's heartbeat (keep-alive). Requires the runId; a superseded runId is a CONFLICT. A run whose heartbeat lapses is swept to EXPIRED and its job to BLOCKED.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "jobId": {
             "minLength": 1,
@@ -9164,7 +9072,6 @@
       },
       "description": "Import contacts from parsed rows (admin scope + ADMIN role). Idempotent by externalId, then lower(email): a matching row UPDATES the existing contact, so a re-run yields zero duplicates. Returns created/updated counts.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "rows": {
             "items": {
@@ -9253,7 +9160,6 @@
       },
       "description": "Import a portable `jentrix.harness-template` document as a NEW template in this workspace (ADMIN; admin scope). It always lands as a DRAFT, never published: the document is untrusted input from another workspace and its credential aliases are unbound, so publishing is a decision you make after reading the returned validation summary. `contentHashMatches` reports whether the document's stated hash matched its definition. idempotencyKey is REQUIRED.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "document": {
             "additionalProperties": {},
@@ -9298,7 +9204,6 @@
       },
       "description": "Import tasks into a column from parsed rows (admin scope + ADMIN role), chunked through the bulk-create op. skipExistingTitles (default true) makes a re-run idempotent — rows whose title already exists on the board are skipped. Returns created/skippedExisting counts.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "columnId": {
             "minLength": 1,
@@ -9384,7 +9289,6 @@
       },
       "description": "Import a Trello board export into a NEW board (admin scope + ADMIN role): lists become columns and cards become tasks. Returns the new board slug and column/card counts.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "idempotencyKey": {
             "description": "Optional idempotency key: retrying with the same key and arguments within 24h returns the original result instead of creating duplicates. Reusing a key with different arguments fails with CONFLICT.",
@@ -9473,7 +9377,6 @@
       },
       "description": "Invite a member to a workspace by email (admin scope; ADMIN role, but an OWNER-role invite requires the OWNER role). The default membership-governance pack parks this behind REQUIRE_APPROVAL, so an agent-alone call surfaces as a human-approvable request; an approved binding lets it through. Acceptance stays with the invitee.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "email": {
             "format": "email",
@@ -9518,7 +9421,6 @@
       },
       "description": "Launch a PUBLISHED harness template version as a new harness run (ADMIN; admin scope), freezing the resolution the run executes under — resolved workers and releases, permission revisions, policies, credential bindings, budgets, gates, repository and work item — verbatim into the run, so what the run did is always answerable. Repo/work-item fields are launch-time and fall back to the version's defaults. Pass launchMode: 'dry_run' for an INERT run: identical rows plus a linked SimulationRun, refused at the claim so no execution leg ever opens, and dry-run evidence can never reach acceptance. Optional expectedChecksum binds the launch to a resolution you already read from preview_harness_template_launch — if it moved since (a re-pointed release, a republished permission revision, an edited policy) the launch fails with CONFLICT instead of freezing something unreviewed. idempotencyKey is REQUIRED. A DRAFT version is never launchable.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "allowedPaths": {
             "items": {
@@ -9607,7 +9509,6 @@
       },
       "description": "Link a credential to an agent (admin scope + ADMIN role): an existing ApiToken owned by a workspace member (workspace-pinned or the caller's own), or an OAuth client for M4 brokered grants. A disabled/deprecated agent is refused. Returns the link id.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "agentId": {
             "description": "Agent to link the credential to.",
@@ -9657,7 +9558,6 @@
       },
       "description": "Link or unlink a workspace contact to a SUPPORT task's ticket. Pass contactId=null to unlink. The contact must belong to the same workspace as the task. (Contacts are the single party directory — use list_contacts / create_contact to find or add people.)",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "contactId": {
             "anyOf": [
@@ -9695,7 +9595,6 @@
       },
       "description": "Link a GitHub repository (from the connected App installation) to a board so its issues sync two-way (admin scope + ADMIN role). repoId is the numeric GitHub repo id as a string; defaultColumnId is where new issues land.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "boardId": {
             "description": "Target board for this repo's synced issues.",
@@ -9780,7 +9679,6 @@
       },
       "description": "Link an M1 Decision to a harness run (admin scope + ADMIN role), optionally scoped to a stage and optionally BLOCKING (the run/stage pauses until the decision is accepted). The decision must be in the run's workspace. Re-linking is an update — pass expectedUpdatedAt to guard against a concurrent change (CONFLICT embeds the current link).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "blocking": {
             "description": "Whether this decision blocks the run/stage until resolved.",
@@ -9843,7 +9741,6 @@
       },
       "description": "Link an existing task to a GitHub issue by number in a connected repo (write scope). Idempotent: a task already linked returns its existing link. The issue then syncs two-way per the repo's config.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "idempotencyKey": {
             "description": "Optional idempotency key: retrying with the same key and arguments within 24h returns the original result instead of creating duplicates. Reusing a key with different arguments fails with CONFLICT.",
@@ -9887,7 +9784,6 @@
       },
       "description": "Adopt an ORPHAN pull-request link onto a task (write scope). The recovery path for a PR whose branch/title/body carried no resolvable task key: sets the owner and flips linkState to MATCHED. The task must be on the board linked to the PR's repo (repo ↔ board is 1:1). Idempotent; pass expectedUpdatedAt for an optimistic-concurrency guard.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "expectedUpdatedAt": {
             "description": "Optimistic concurrency guard: the PR link's updatedAt from your last read. If it changed since, the call fails with CONFLICT and the error embeds the link's current state.",
@@ -9928,7 +9824,6 @@
       },
       "description": "List the activity log for a task or a whole board, newest first. Filter by since (ISO timestamp), source ('ui' = humans, 'mcp' = agents), or activity type — e.g. diff what humans did since your last run.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "boardId": {
             "description": "Scope to a whole board. Provide taskId or boardId.",
@@ -10174,7 +10069,6 @@
       },
       "description": "List the workspace's agent-workflow jobs (the plan→review→execute→review loops), filterable by phase and task. Poll get_agent_job for full state.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "anyOf": [
@@ -10237,7 +10131,6 @@
       },
       "description": "List an agent's immutable releases (content-addressed configuration pins). Filter by agentProfileId or status (CANDIDATE/ELIGIBLE/RETIRED). Each row carries its releaseNo, configurationDigest, model, gating eval run + dataset version, and isActive — 'active' is derived SOLELY from the profile's pointer, never a status. Cursor-paginated; call get_agent_release for the full pins + diff-vs-prior + eval evidence + pointer history.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "agentProfileId": {
             "minLength": 1,
@@ -10275,7 +10168,6 @@
       },
       "description": "List connected sessions in a workspace, filterable by project, operator, provider, lifecycle status, and captureComplete (the two dimensions are independent). Paginated; provider thread ids are omitted from list output.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "captureComplete": {
             "type": "boolean"
@@ -10336,7 +10228,6 @@
       },
       "description": "List the workspace's registered agents (teammate identities) with role, provider, runtime, trust level, status, latest eval score, recent failures, routing boundaries, and credential count. Filter by status/role. Paginated; pass nextCursor from the previous response to continue.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "anyOf": [
@@ -10403,7 +10294,6 @@
       },
       "description": "List first-class artifacts (PR/diff/log/screenshot/eval report/run summary/…) in a workspace, filterable by project/task/run/workOrder/session/type/creator. Internal (R2) artifacts expose hasObject; get_artifact returns the signed download URL. Redacted artifacts remain as tombstones.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "creatorId": {
             "minLength": 1,
@@ -10486,7 +10376,6 @@
       },
       "description": "List a task's attachments with short-lived signed download URLs (~5 min). Uploads are UI-only.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "taskId": {
             "description": "Task whose attachments to list.",
@@ -10509,7 +10398,6 @@
       },
       "description": "List a workspace's automation rules (admin scope), optionally filtered to one board plus workspace-wide rules. Shows trigger/conditions/actions, run counts, and enabled state.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "boardId": {
             "description": "Filter to rules scoped to this board (plus workspace-wide ones).",
@@ -10536,7 +10424,6 @@
       },
       "description": "List boards in a workspace.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "workspaceId": {
             "description": "Workspace ID. Discover via list_workspaces.",
@@ -10559,7 +10446,6 @@
       },
       "description": "List CI runs in a workspace (newest first) with status, conclusion, branch, and check roll-up counts. Filter to one PR (prLinkId) or one task (taskId). Paginated — pass nextCursor to continue.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "anyOf": [
@@ -10610,7 +10496,6 @@
       },
       "description": "List columns of a board, ordered by position. Columns represent statuses on the kanban board.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "boardId": {
             "description": "Board ID. Discover via list_boards(workspaceId).",
@@ -10633,7 +10518,6 @@
       },
       "description": "List a task's comments oldest-first, with authors.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "taskId": {
             "description": "Task whose comments to list.",
@@ -10656,7 +10540,6 @@
       },
       "description": "List governed communication connectors (Slack/email/calendar/doc). Each extends a source record carrying trust/freshness/privacy/owner/project-scope; outboundEnabled gates whether drafts may target it.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "workspaceId": {
             "minLength": 1,
@@ -10678,7 +10561,6 @@
       },
       "description": "List/search workspace contacts, optionally filtered by query (substring match on name/email/company), with total count. Concise by default — pass response_format: 'detailed' for phone/externalId/timestamps. Paginated; pass nextCursor from the previous response to continue.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "anyOf": [
@@ -10737,7 +10619,6 @@
       },
       "description": "List the workspace's OPEN, actionable Control-Tower items across agent workflow, delivery, harness, SLA, eval, policy, and credential queues, ranked highest-severity then oldest. Governed memory review uses the dedicated Memory page; the legacy MEMORY_CANDIDATES adapter has no MemoryItem source. Each item carries a queue, stable sourceKey, title, an 'explain why this is here' reason, severity, optional project/agent, first-seen time, and valid actions. Paginated; pass nextCursor to continue.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "agentProfileId": {
             "description": "Filter to one agent.",
@@ -10823,7 +10704,6 @@
       },
       "description": "List Decision Register entries in a workspace with owner, status, lineage, memory status, and revision counts. Filter by project, status, or memory status. Paginated.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "anyOf": [
@@ -10888,7 +10768,6 @@
       },
       "description": "List the delivery graph's pull-request links in a workspace (newest first), each with PR state, link state (MATCHED/ORPHAN), owning task key, and aggregate CI signal. Filter by board, linkState (ORPHAN = adoptable PRs), or PR state. Paginated — pass nextCursor to continue.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "boardId": {
             "description": "Filter to one board's linked repo.",
@@ -10951,7 +10830,6 @@
       },
       "description": "List eval cases in a workspace (fixture input + grader reference + origin/tags). Filter by origin (manual/incident/mined) or graderKind (deterministic/model_graded/trace). Cursor-paginated. Call get_eval_case for the full fixture + grader config.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "minLength": 1,
@@ -10993,7 +10871,6 @@
       },
       "description": "List eval datasets in a workspace (stable identity + mutable head pointer). Each row carries its key, head version id, and version count. Cursor-paginated. Call get_eval_dataset for its immutable version list.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "minLength": 1,
@@ -11019,7 +10896,6 @@
       },
       "description": "List eval runs in a workspace (dataset version × subject × trigger). Filter by subjectKind/subjectId, status, datasetId, or gating. Cursor-paginated concise rows — call get_eval_run for the per-case results + clickable traces.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "minLength": 1,
@@ -11077,7 +10953,6 @@
       },
       "description": "List shadow/canary experiments in a workspace (candidate-vs-baseline comparisons over the release pointer). Filter by agentProfileId, status (RUNNING/PROMOTED/ROLLED_BACK/SUPERSEDED/STALE/EXPIRED), or mode (SHADOW/CANARY). Each row carries its experimentNo, baseline/candidate release, traffic fraction, and window. A RUNNING canary is a deterministic routing override — the pointer stays on the baseline. Cursor-paginated; call get_experiment for the guardrail spec + governance snapshot + comparison report.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "agentProfileId": {
             "minLength": 1,
@@ -11123,7 +10998,6 @@
       },
       "description": "List the workspace's GitHub App connection and how many repos are linked (admin scope). Empty when GitHub isn't connected.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "workspaceId": {
             "description": "Workspace ID.",
@@ -11146,7 +11020,6 @@
       },
       "description": "List tasks linked to GitHub issues in a workspace with the issue url, state, and last-synced time (newest first, max 100).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "workspaceId": {
             "description": "Workspace ID.",
@@ -11169,7 +11042,6 @@
       },
       "description": "List a harness run's review findings (severity, category, status, title, the stage they target). Filter by stage/status/severity. Concise by default — pass response_format: 'detailed' for the body, reviewer, resolution note, verification refs, and the dedup identity (fingerprint, reopenCount, basisKind/basisRef): findings sharing a fingerprint are ONE defect reconciled across rounds, so re-raising one bumps reopenCount instead of filing a new row. Paginated; pass nextCursor to continue.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "anyOf": [
@@ -11250,7 +11122,6 @@
       },
       "description": "List a harness run's stages in order: status, the human gate it parks at, branch, the next build action, the open-finding + evidence (artifact) counts, and launchable (whether the stage has a backing Work Order — false for a draft stage; it means 'not a draft', NOT 'launchable right now', since serial ordering + run/stage status still apply at launch). Filter by status. Concise by default — pass response_format: 'detailed' for Work Order/lease ids, commit policy, acceptance criteria, and timestamps. Paginated; pass nextCursor to continue.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "anyOf": [
@@ -11316,7 +11187,6 @@
       },
       "description": "List a workspace's harness templates — reusable, versioned pipeline definitions a run is launched from (ADMIN-authored, MEMBER-readable). Each row carries the identity, the draft/published version numbers, the usage rollup, and the draft's validation COUNTS (pass the template to validate_harness_template for the findings behind them). Concise by construction: no definition documents. `updatedAt` is the value you pass back as expectedUpdatedAt on the next mutation. Ordered most-recently-updated first; a truncated page says so in `notice`.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "includeArchived": {
             "type": "boolean"
@@ -11347,7 +11217,6 @@
       },
       "description": "List Harness Automation runs in a workspace with status, current stage, and a live done/total stage count. Filter by project/task/status. Concise by default — pass response_format: 'detailed' for repo/branch config, freeze state, and timestamps. Paginated; pass nextCursor to continue.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "anyOf": [
@@ -11416,7 +11285,6 @@
       },
       "description": "List incidents (failed runs / bad actions escalated into postmortems). Open-first by default; filter by status/severity/owner. Concise rows — call get_incident for the full timeline + linked entities.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "minLength": 1,
@@ -11464,7 +11332,6 @@
       },
       "description": "List workspace initiatives (across all ROADMAP boards), with their parent board and total count. Paginated.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "anyOf": [
@@ -11511,7 +11378,6 @@
       },
       "description": "List repositories the connected GitHub App installation can access (admin scope + ADMIN role). Each row carries the repo id (as a string), owner/name, private flag, and whether it is already linked to a board (a repo links to exactly one board). Refused with a structured INVALID_INPUT when GitHub is not connected.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "workspaceId": {
             "description": "Workspace ID.",
@@ -11534,7 +11400,6 @@
       },
       "description": "List a workspace's pending (unaccepted) invitations (admin scope + ADMIN role): id, email, role, inviter, created/expiry. The invite token is never returned. Use the id with revoke_invitation.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "workspaceId": {
             "minLength": 1,
@@ -11556,7 +11421,6 @@
       },
       "description": "List labels available on a board.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "boardId": {
             "description": "Board ID.",
@@ -11579,7 +11443,6 @@
       },
       "description": "List work leases (coordination claims) in a workspace, newest first. Filter by targetType/status/holderAgentId. A lease says an actor is actively working a target for a bounded window. Check this before launching agent work to avoid duplicate/colliding runs.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "type": "string"
@@ -11632,7 +11495,6 @@
       },
       "description": "List ingested stakeholder messages/meetings linked to an entity (task/project/contact/decision/Work Order) — the linked-conversation panel + meeting commitments.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "contactId": {
             "anyOf": [
@@ -11713,7 +11575,6 @@
       },
       "description": "List members of a workspace. Use the userId field when assigning tasks.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "workspaceId": {
             "description": "Workspace ID.",
@@ -11736,7 +11597,6 @@
       },
       "description": "List a workspace's governed memory items (provenance-anchored units of durable knowledge). Each row carries its kind, lifecycle status, provenance (source kind + id), and `selectable` — true ONLY for ACCEPTED_TO_MEMORY items, the only status ever injected into a launch manifest. NOMINATED (proposed) and QUARANTINED (source superseded; parked for human re-review) items are visible but not selectable. Filter by status/sourceKind; cursor-paginated, newest first.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "minLength": 1,
@@ -11790,7 +11650,6 @@
       },
       "description": "List recorded outcome metrics (cycle time, reopen rate, human edits, approval burden, delivery metrics, …) for a workspace, comparing agent-assisted vs human-only cohorts where applicable. substrateGated=true marks legacy metric slots not yet derived from the newer run/eval registries. Filter by kind/cohort/project/agent.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "agentId": {
             "minLength": 1,
@@ -11856,7 +11715,6 @@
       },
       "description": "List promoted playbooks (reusable successful workflows) with their trigger, required policy/eval coverage, and enabled state.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "minLength": 1,
@@ -11882,7 +11740,6 @@
       },
       "description": "List the workspace's layered policies (system/workspace/project/agent/work-order/grant) with their template, governed domain, outcome, priority, and enabled state. Filter by domain/scope. Reads never surface credential backend references or secrets.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "domain": {
             "enum": [
@@ -11929,7 +11786,6 @@
       },
       "description": "List the workspace's policy evaluation feed (append-only), newest first, with each action's outcome and reason. Filter by runId/outcome/domain/since. Paginated; pass nextCursor to continue. Denials also surface in the Control Tower policy queue.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "type": "string"
@@ -11988,7 +11844,6 @@
       },
       "description": "List projects in a workspace with owner, status, and child/link/milestone/risk/decision counts. Filter by status, owner, or parent project. Paginated; pass nextCursor from the previous response to continue.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "anyOf": [
@@ -12060,7 +11915,6 @@
       },
       "description": "List a workspace's provider connections (ADMIN). A provider connection is how a registered runtime adapter (claude/codex) bills and where it sends requests. Each row is redacted at the ops-core boundary — a connection references its secret by id only (credentialRecordId), never any secret material — and carries its verification state (unverified / verified / failed / configured_but_unverified), the per-adapter verification source (server probe vs runner attestation), and, for subscription connections, the derived 'verified by runner NAME at TIME' summary plus the list of nodes currently eligible. Use it to see which connections are healthy and where a subscription can run.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "includeArchived": {
             "type": "boolean"
@@ -12085,7 +11939,6 @@
       },
       "description": "List an agent's release-pointer transitions (the activation audit trail), newest first. Filter by agentProfileId or toReleaseId. Each transition IS the immutable activation-verdict snapshot — reason (activation/promotion/rollback/waiver_activation), cited eval run + evidence digest, configuration digest, dataset version + observed head, policy revision pins, approver, and any waiver — so every pointer move is a visible, self-describing audit event. Cursor-paginated.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "agentProfileId": {
             "minLength": 1,
@@ -12119,7 +11972,6 @@
       },
       "description": "List dry-run simulations (proposed/approved/executed/expired). Concise rows — call get_simulation for the staged plan + projected side effects.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "minLength": 1,
@@ -12154,7 +12006,6 @@
       },
       "description": "List a workspace's skills (reusable, versioned agent capabilities). Each row carries its stable key, name, head version number + status (CANDIDATE/ELIGIBLE/RETIRED), and total version count — 'head' is the mutable current-version pointer; the versions themselves are immutable and content-addressed. Cursor-paginated; call get_skill for the full version history + executable snapshots.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "minLength": 1,
@@ -12181,7 +12032,6 @@
       },
       "description": "List governed knowledge sources (trust/freshness/privacy/owner/client/supersession). Superseded sources are EXCLUDED by default — pass includeSuperseded to show them. Client-private sources are blocked from unauthorized cross-client contexts.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "type": "string"
@@ -12230,7 +12080,6 @@
       },
       "description": "List executable specs (user story / constraints / non-goals / test plan / rollback / evidence / definition of done) in a workspace. Filter by taskId/projectId/status. Accepted specs feed Work Orders.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "type": "string"
@@ -12272,7 +12121,6 @@
       },
       "description": "List a task's outgoing and incoming links with the linked tasks' keys, titles, and boards.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "taskId": {
             "description": "Task whose outgoing and incoming links to list.",
@@ -12295,7 +12143,6 @@
       },
       "description": "List tasks in a board or column with board/column names, total count, and pagination. Concise by default — pass response_format: 'detailed' for sidecars/labels/assignees. At least one of boardId or columnId is required.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "boardId": {
             "description": "Filter by board. Provide either boardId or columnId.",
@@ -12350,7 +12197,6 @@
       },
       "description": "List a webhook's recent deliveries with status/attempts/errors (admin scope), newest first. Paginated.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "anyOf": [
@@ -12391,7 +12237,6 @@
       },
       "description": "List a workspace's outbound webhooks (admin scope). Secrets are never returned; consecutiveFailures > 0 means recent deliveries are exhausting their retries (3 in a row auto-disables the webhook).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "workspaceId": {
             "description": "Workspace ID.",
@@ -12414,7 +12259,6 @@
       },
       "description": "List Work Orders in a workspace with task key, status, current version number, and version/launch counts. Filter by task, project, or status. Paginated; pass nextCursor from the previous response to continue.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "anyOf": [
@@ -12478,7 +12322,6 @@
       },
       "description": "List the managed-execution workers enrolled in this deployment (ADMIN). Workers are deployment-global platform resources — not tenant data — so this shows identity, driver class + whether it is an approved production class, image/app/schema revision, liveness (live/stale/disabled via the 30s heartbeat cadence), and a COUNT of the worker's current claims. Current claims are derived from run rows at read time; no per-workspace claim detail is exposed. Use it to see which workers can accept managed eval work and which are down.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "includeDisabled": {
             "type": "boolean"
@@ -12503,7 +12346,6 @@
       },
       "description": "List workspaces the authenticated user belongs to, with their role.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {},
         "type": "object"
       },
@@ -12519,7 +12361,6 @@
       },
       "description": "Create, rename, move, or archive a board column (action: 'create' | 'rename' | 'move' | 'archive'). create needs boardId+name; rename needs columnId+name; move needs columnId + beforeColumnId/afterColumnId; archive needs columnId (tasks in it disappear from the board with it).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "action": {
             "description": "Which column operation to perform.",
@@ -12589,7 +12430,6 @@
       },
       "description": "Create, update, or delete a board label (action: 'create' | 'update' | 'delete'). create needs boardId+name+color; update needs labelId and name and/or color; delete needs labelId. Creating an existing name returns the existing label. Use set_task_labels to apply labels to tasks.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "action": {
             "description": "Which label operation to perform.",
@@ -12639,7 +12479,6 @@
       },
       "description": "Map a GitHub user (numeric id as a string + login) to a Jentrix workspace member, or pass userId: null to unmap (admin scope + ADMIN role). The target must be a workspace member. Upsert — idempotent.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "githubLogin": {
             "minLength": 1,
@@ -12687,7 +12526,6 @@
       },
       "description": "Mark a CANDIDATE release ELIGIBLE from a passing gating eval run (admin scope). The run must be a gating AGENT_RELEASE run whose status is PASSED, whose configuration digest equals the release's, and (when the release pins one) whose dataset version matches the pinned gating dataset version — the M13.4 evidence bar, verified server-side. One-way: ELIGIBLE never reverts (retire instead); marking eligible records the verdict but grants no authority — set_active_release is the separate new-authority activation. AGE-248: this transition previously had no product surface, deadlocking managed claims on fresh deployments. Returns the release detail row.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "gatingEvalRunId": {
             "minLength": 1,
@@ -12716,7 +12554,6 @@
       },
       "description": "Merge one contact into another (admin scope) — the duplicate's deal links and tickets move to the survivor, empty survivor fields are filled from the duplicate, and the duplicate row is DELETED, all in one transaction. Cannot merge a contact into itself.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "duplicateId": {
             "description": "Contact folded into the survivor and deleted.",
@@ -12747,7 +12584,6 @@
       },
       "description": "Move a task to a different column (= change status on the kanban board) and/or reorder it. Use beforeTaskId/afterTaskId for placement; omit both to append to the end.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "afterTaskId": {
             "anyOf": [
@@ -12806,7 +12642,6 @@
       },
       "description": "Resolve what launching a published template version WOULD freeze, without writing anything (ADMIN role; read scope) — the launch review, as a tool. Returns the full resolution (workers and the immutable releases they pin, compiled permission revisions, policy revisions, credential bindings, budgets, gates, repository and work item), its `checksum`, and everything standing in the way: `blockers` refuse the launch, `warnings` do not, and `missingPrerequisites` says it in a sentence. Pass the returned checksum to launch_harness_template as expectedChecksum to bind the launch to exactly this resolution — if it moved since (a re-pointed release, a republished permission revision, an edited policy) the launch CONFLICTs instead of freezing something you never read. Takes every argument the launch takes, because a preview that resolved different inputs would be a preview of a different run.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "allowedPaths": {
             "items": {
@@ -12882,7 +12717,6 @@
       },
       "description": "Promote a healthy experiment candidate (admin scope + ADMIN role) — the explicit human action that moves the release pointer EXACTLY once. Writes the ReleasePointerTransition to the candidate AND transitions the experiment RUNNING → PROMOTED in ONE transaction, re-verifying the activation currency checks (digest-equal green gating run, current policy pins, dataset version at the head). Guardrail rollback, supersession (an admin moved the pointer mid-window), and staleness (governance drift) are the AUTOMATIC terminal paths and have no MCP surface — promotion is the only human, pointer-moving one. Use expectedRevision / expectedPointerRevision for safe retries. Returns the pointer transition it wrote.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "approvalRef": {
             "minLength": 1,
@@ -12928,7 +12762,6 @@
       },
       "description": "Mint a DRAFT harness template from a run that already worked (ADMIN; admin scope) — its stages become the steps, its roles the assignments, its repo/branch settings the defaults. Eligibility is strict and fails closed: the run must be unarchived, COMPLETED, and acceptanceState ACCEPTED. PENDING and WAITING_OPERATOR are verdicts still in flight; CONDITIONAL is a waiver that must never read as verified; LEGACY_UNATTESTED means acceptance was never established; a DRY RUN never promotes because nothing in it executed — each is refused with the reason. A SKIPPED stage promotes to an ARCHIVED step, so the pipeline's shape survives without resurrecting skipped work. Promotion buys provenance, never a validation bypass: the draft carries an ordinary validation summary and still has to pass publish. idempotencyKey is REQUIRED.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "harnessId": {
             "description": "Alias for harnessRunId. Accepted interchangeably — pass either one; if both are given they must be equal.",
@@ -12969,7 +12802,6 @@
       },
       "description": "Publish a harness template's open draft, freezing it as an immutable version that launches can name (ADMIN; admin scope). VALIDATION GATES THIS: a draft carrying any `error` finding is refused (warnings never block) — call validate_harness_template first to see what would refuse. A template with no open draft is refused too; edit it to fork one. expectedUpdatedAt is REQUIRED. Returns the published version's id, number, and content hash.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "expectedUpdatedAt": {
             "description": "REQUIRED on this surface: the template's `updatedAt` from your last read (ISO 8601). If the template changed since, the call fails with CONFLICT and the error embeds the current state under error.current so you can merge without a second read.",
@@ -13000,7 +12832,6 @@
       },
       "description": "Publish a worker's permission revision (admin scope + ADMIN role) — the boundary that GRANTS an AI worker its compiled permissions. Compilation is atomic: an uncompilable content is refused and nothing is written, so a returned revisionId proves the compiled enforcement bundle exists and is current. This is what clears the permission_revision_unset and compiled-artifact staleness blockers a launch preview reports. Governed by the worker.permission.publish policy key (seeded REQUIRE_APPROVAL), so a human approves the grant even though an agent can request it.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "agentProfileId": {
             "description": "The workspace AI worker whose permission template to publish into.",
@@ -13164,7 +12995,6 @@
       },
       "description": "Run the workspace's software-company reconciler once, on demand (ADMIN). This is the SAME idempotent operation the five-minute sweep runs for each enabled profile — the manual recovery/catch-up path: it re-derives all pending work from durable rows (signal intake, triage, investment-decision materialization, harness handoffs, CI-failure repair, release/deploy verification, incident recovery, support response, company-operations approvals, and the weekly operating review), advances what is ready, opens COMPANY_AUTOMATION exceptions for blocked contracts, and clears exceptions a prior retry has resolved (audit history preserved). Returns the created / advanced / skipped / failed subjects plus the count of cleared exceptions. A paused or disabled profile starts no new work (empty result). Never overrides freeze, cancellation, gate, lease, connector-disable, rollback, or release controls — it only invokes the existing operations as the configured operator.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "workspaceId": {
             "minLength": 1,
@@ -13188,7 +13018,6 @@
       },
       "description": "Record that a human performed an OPERATOR acceptance criterion of a harness run's acceptance policy (deploy, env restamp, runner restart, drill …). The criterion must match the policy verbatim and be OPERATOR mode — AUTOMATED criteria are proven by runner-attested verification commands and are rejected here. Captures a run-scoped evidence artifact (note redacted on ingest, criterion recorded on the link), writes the audit activity, and recalculates the run's acceptanceState in the same flow. Returns the new acceptanceState. Admin only.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "artifactId": {
             "anyOf": [
@@ -13256,7 +13085,6 @@
       },
       "description": "Record a §10.6 next action on an execution (admin scope + ADMIN role), optionally with the TERMINAL DISPOSITION of a failed or cancelled one. The action is REFUSED unless legal for the run's current state; a failed/cancelled execution carrying no disposition must be given one. The disposition columns and the append-only history row are written in one transaction. expectedUpdatedAt guards the write (CONFLICT embeds the current state). Returns the full resolution view: legalActions, disposition, and history.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "action": {
             "description": "The §10.6 action taken; refused unless legal for the state.",
@@ -13342,7 +13170,6 @@
       },
       "description": "Runner only: attach provider-reported token/cost usage to an attempt run AFTER its turn ended (legal on a closed run). Write-once — a run that already carries usage returns recorded=false. inputTokens is TOTAL input; cacheReadTokens/cacheCreationTokens are disjoint subsets of it. Never estimate: omit what the runtime did not report.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "jobId": {
             "minLength": 1,
@@ -13450,7 +13277,6 @@
       },
       "description": "Release a work lease you hold (status → released). Idempotent: releasing an already-inactive lease is a no-op. Pass expectedUpdatedAt for an optimistic-concurrency guard.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "expectedUpdatedAt": {
             "description": "Optimistic-concurrency guard: the lease's updatedAt from your last read. If it changed since, the call fails with CONFLICT.",
@@ -13484,7 +13310,6 @@
       },
       "description": "Remove a member from a workspace (admin scope + ADMIN role). You cannot remove YOURSELF, and the last OWNER cannot be removed. Governed by the membership pack (REQUIRE_APPROVAL by default). Concurrency is VALUE-CAS: pass `expectedRole`; `expectedUpdatedAt` is refused (M19.2). A value-CAS mismatch returns CONFLICT embedding the current role.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "expectedRole": {
             "description": "Value-CAS guard: the member's role you last read. If it changed since, the call fails with CONFLICT.",
@@ -13526,7 +13351,6 @@
       },
       "description": "Remove a project link (exact projectId + targetType + targetId match). Removing a link that does not exist is a no-op. Needs the workspace ADMIN role.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "projectId": {
             "minLength": 1,
@@ -13569,7 +13393,6 @@
       },
       "description": "Remove a task link (exact fromTaskId + toTaskId + kind match). No-op if the edge does not exist.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "fromTaskId": {
             "description": "Task ID on the FROM side.",
@@ -13611,7 +13434,6 @@
       },
       "description": "Rename a board.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "boardId": {
             "description": "Board to rename.",
@@ -13643,7 +13465,6 @@
       },
       "description": "Start a review-loop workflow on a task: creates the job + a Work Order draft (no launch yet) and names the planner/reviewer agents. The planner drafts, the reviewer reviews, with human gates before execution and at final acceptance. The `authoring` template is the exception: it converges on a validated spec, so approving its plan (authoring) gate completes the loop — there is no execution phase or final-acceptance gate.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "executorAgentId": {
             "description": "AgentProfile that executes. v1: must equal plannerAgentId (the planner executes) — a distinct executor is not yet routable.",
@@ -13717,7 +13538,6 @@
       },
       "description": "Mint a short-lived, scoped, action-bound credential grant from a credential record (admin only, policy-gated). The grant material is minted only at the injection boundary — never a long-lived raw secret.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "actionClass": {
             "description": "Exact action class the grant authorizes, e.g. repo:push.",
@@ -13783,7 +13603,6 @@
       },
       "description": "Start a delivery workflow on a task (feature | bugfix | ci_failure_triage | release_candidate | hotfix), optionally bound to an existing PR (prLinkId, on the task's own board). Same plan→review→execute→review loop as request_agent_workflow, but the reviewer turn reacts to the bound PR's CI (see get_agent_job's `delivery` block) and the ci_failure_triage executor pushes a fix to the PR branch via the governed push path. Merge/deploy stay human-gated (M8.3).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "idempotencyKey": {
             "maxLength": 200,
@@ -13858,7 +13677,6 @@
       },
       "description": "Request execution of the run's first undone stage (execution is serial, so it is the only one in line): acquires the stage lease + opens a fresh M7 AgentJob bound to the stage. The stage must have executable backing — a first-undone draft stage (no Work Order, e.g. a create_harness_stage draftOnly placeholder) is rejected as a draft, never launched. Optionally pass stageId to assert which stage you expect to launch — a mismatch (a different stage is next) is rejected rather than silently launching the wrong one. Long-running: task-augmentable — cancelling the task (tasks/cancel) cancels the underlying stage-execution kernel workload. Returns whether a stage was launched + the stage/job ids.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "harnessId": {
             "description": "Alias for harnessRunId. Accepted interchangeably — pass either one; if both are given they must be equal.",
@@ -13904,7 +13722,6 @@
       },
       "description": "Request an agent run for a Work Order. With no live executor wired, the launch is recorded in PENDING for a runner to claim (substrate fallback) and a notice is returned. Pass quickRun for a restricted-tools quick launch.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "allowedTools": {
             "description": "Restricted tool set for a quick run.",
@@ -13968,7 +13785,6 @@
       },
       "description": "Resolve a require_approval PolicyEvaluation by minting a scoped, time-bounded approval binding (admin only). The approved actor may then retry the gated action once within the binding's window.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "evaluationId": {
             "description": "The PolicyEvaluation (require_approval) being resolved.",
@@ -14008,7 +13824,6 @@
       },
       "description": "Resolve non-archived projects across every workspace the caller belongs to whose linked repository matches a normalized \"owner/name\". Used by connected-session setup to narrow project choices from the checkout; requires an unpinned human credential (workspace-pinned tokens are refused).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "repoOwnerName": {
             "description": "Normalized repository identity as lowercase \"owner/name\" (e.g. \"acme/api\").",
@@ -14035,7 +13850,6 @@
       },
       "description": "Reopen an eligible INTERRUPTED session (original operator or workspace admin; same project and repository — continuity failures are SESSION_RESUME_CONFLICT). Requires expectedUpdatedAt (CAS). A deliberate new work period should instead create a new session with continuesSessionId.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "expectedUpdatedAt": {
             "description": "REQUIRED optimistic concurrency guard: the session's updatedAt from your last read; a mismatch fails with CONFLICT embedding the current state.",
@@ -14080,7 +13894,6 @@
       },
       "description": "Retire a release (admin scope + ADMIN role) — a terminal move from CANDIDATE/ELIGIBLE to RETIRED. Idempotent: retiring an already-RETIRED release replays it. REFUSED with CONFLICT when the pointer currently holds the release (activate or roll back to another first). Returns the retired release.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "releaseId": {
             "minLength": 1,
@@ -14104,7 +13917,6 @@
       },
       "description": "Retry a stage parked BLOCKED by a FAILED or stalled execution job (runner crash, provider error, lapsed heartbeat) — moves it BLOCKED → READY so it becomes relaunchable, then request_harness_stage_run opens a fresh attempt. Must be the run's current (first-undone) stage on an EXECUTING run. NOT for a stage blocked by an unresolved decision (accept the decision and resume the stage instead — this is rejected to avoid bypassing the decision gate). Admin only.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "harnessId": {
             "description": "Alias for harnessRunId. Accepted interchangeably — pass either one; if both are given they must be equal.",
@@ -14145,7 +13957,6 @@
       },
       "description": "Revoke a pending workspace invitation (admin scope + ADMIN role). Discover the invitationId via list_invitations.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "invitationId": {
             "minLength": 1,
@@ -14169,7 +13980,6 @@
       },
       "description": "Run an eval: grade a subject configuration (AGENT_CONFIG / PLAYBOOK / HARNESS_TEMPLATE) against a dataset version (explicit datasetVersionId, else the dataset's head) on the managed kernel, then land a terminal EvalRun + per-case results via the outbox. A per-run budget is MANDATORY — set at least one of maxCases/maxTokens/maxSpendUsd. Pass gating:true to export unsampled and capture content-addressed evidence (a gating verdict gates promotion; running an eval never grants authority). Long-running: task-augmentable (tasks/get for status, tasks/cancel to cancel the kernel workload); the run id + terminal status/score come back either way, and the per-case results read from get_eval_run. The gating configuration digest is derived SERVER-SIDE — from the profile for AGENT_CONFIG and from the draft + coverage for PLAYBOOK (never caller-supplied); only HARNESS_TEMPLATE still takes configurationDigest. PLAYBOOK gating runs are driven from the promotion flow, not this tool.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "configurationDigest": {
             "description": "Only for HARNESS_TEMPLATE subjects (ignored for AGENT_CONFIG/PLAYBOOK — those digests are derived server-side).",
@@ -14279,7 +14089,6 @@
       },
       "description": "Save (upsert) an AI worker draft on a workspace agent (admin scope + ADMIN role) — the configure→publish flow's persisted snapshot: runtime adapter + provider connection + model profile + generation controls + optional fallback. The triple is validated server-side (an unsupported control or a connection/model that doesn't belong to the adapter is refused). expectedUpdatedAt optionally guards against a concurrent profile edit. Returns the resolved draft.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "agentId": {
             "minLength": 1,
@@ -14399,7 +14208,6 @@
       },
       "description": "Search the BODIES of typed artifacts (plan/decision/findings/report/diff/deliverable/learning/prompt/goal/prd) workspace-wide — full-text, websearch syntax (words, quoted phrases, -negation). This is how a LEARNING or a FINDINGS write-up is found again by what it says rather than by a title someone remembers. The index is built from the REDACTED stored text, so a secret redacted at ingest was never indexed. Archived and redacted artifacts are excluded; raw TRACE transcript parts are not searchable. Fetch a hit's content with get_artifact.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "query": {
             "description": "Search text matched against artifact titles and their stored (redacted) bodies. Websearch syntax supported.",
@@ -14435,7 +14243,6 @@
       },
       "description": "Search tasks workspace-wide across all boards: exact task key (for example STK-42), full-text (websearch syntax — words, quoted phrases, -negation) over title/description/comments, plus case-insensitive substring on title/description. Exact key hits rank first. Returns concise rows with board/column names. Archived tasks excluded.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "cursor": {
             "anyOf": [
@@ -14483,7 +14290,6 @@
       },
       "description": "Discover this MCP server's own tools: returns capability manifests (name, description, required scope, safety annotations, parameter names/types) filtered by free-text query, scope class, or task profile. Reads the CI-generated, drift-pinned tool registry — not a live introspection — so results are stable and match the frozen surface. Pass include_profiles to also get the registry+telemetry-derived task profiles (recommended tool bundles per workflow). Read-only; no workspace needed.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "include_profiles": {
             "description": "Also return the registry+telemetry-derived task profiles (tool bundles per workflow).",
@@ -14529,7 +14335,6 @@
       },
       "description": "Idempotently seed the starter eval datasets into a workspace (admin scope + ADMIN role): 'platform-smoke' (one case per grader kind), 'platform-smoke-deterministic' (judge-free grader smoke), and 'platform-smoke-managed-release' (the real-provider gate with an explicit JSON response contract). Release gating should pin the returned gatingDatasetVersionId (the managed release set): its predicate grades the actual managed-provider output, never the fixture's golden subject output. Re-running returns existing immutable head versions (created: false). AGE-248: the workspace bring-up affordance — a workspace-governed release must pin a gating dataset version.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "idempotencyKey": {
             "description": "Optional idempotency key: retrying with the same key and arguments within 24h returns the original result instead of creating duplicates. Reusing a key with different arguments fails with CONFLICT.",
@@ -14560,7 +14365,6 @@
       },
       "description": "Send an outbound reply on a governed communication connector (Slack/email), in the context of an ingested stakeholder message. LIVE delivery (M14.3d), FAIL-CLOSED at every layer: the source must have outbound enabled; the Policy Engine must not DENY or park the send; and the PROVENANCE FLOOR applies — a policy ALLOW delivers a human send, but this AGENT-initiated call parks at REQUIRE_APPROVAL (naming its provenance) unless a bound single-use approval for THIS exact payload authorizes it. `delivered:true` is reported ONLY from the provider acknowledgment; a transient failure leaves the send PENDING for the sweep drain to retry. A workspace with no live connector configured behaves byte-identically to v1 (delivered:false, the 'no live connector' notice). Admin scope + ADMIN role — provenance, not privilege, is what the floor gates.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "body": {
             "maxLength": 50000,
@@ -14607,7 +14411,6 @@
       },
       "description": "Move an agent's active-release pointer (admin scope + ADMIN role). mode 'activate' is the new-authority grant — requires a passing gating EvalRun whose subject digest equals the release's, with policy pins current and the pinned dataset version at the regression head (a stale head returns CONFLICT with the new head named; pass a governed waiverId to override). mode 'rollback' re-points to a prior ELIGIBLE release WITHOUT re-running the new-authority gate (the safe direction). Every move writes ONE reason-bearing, CAS-guarded ReleasePointerTransition — the visible activation-verdict audit event. Agents never self-approve: activation is an admin action. Returns the transition it wrote.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "approvalRef": {
             "minLength": 1,
@@ -14664,7 +14467,6 @@
       },
       "description": "Enable or disable an automation rule (admin scope + ADMIN role).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "automationId": {
             "description": "Alias for ruleId. Accepted interchangeably — pass either one; if both are given they must be equal.",
@@ -14697,7 +14499,6 @@
       },
       "description": "Set the full list of contacts attached to a CRM task's deal. REPLACES existing contacts — pass current+new to add, pass current-removed to remove, pass [] to clear. All contacts must belong to the same workspace as the task and not be archived. Writes one CONTACT_LINKED activity per added contact, one CONTACT_UNLINKED per removed.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "contactIds": {
             "description": "Full list of contact IDs to attach to this deal. REPLACES all existing contacts — pass current+new to add one, pass current-removed to remove one, pass [] to clear. Every contact must belong to the same workspace as the task and not be archived.",
@@ -14731,7 +14532,6 @@
       },
       "description": "Set a harness run's active PLANNER and REVIEWER agents so it can launch stage execution jobs. Pass an AgentProfile id per role (plannerAgentId, reviewerAgentId) — each must be an ACTIVE agent usable in the run's workspace with exactly one usable linked PAT; the token is resolved from the agent, never passed in. Supersedes prior role assignments (archived, not deleted, so history stays auditable). Pass expectedUpdatedAt for optimistic concurrency. Admin only.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "expectedUpdatedAt": {
             "description": "Optimistic concurrency guard: the entity's updatedAt from your last read. If it changed since, the call fails with CONFLICT and the error embeds the current state.",
@@ -14779,7 +14579,6 @@
       },
       "description": "Change a harness run's loop policy: budgets (maxIterations), ordering (sequential), the reviewer requirement (reviewerRequired), gate mode (requireHumanGates — default false: an APPROVED review auto-resolves the gate it opened; true parks every gate for a human), stage auto-advance (autoAdvanceStages — default true: the next stage launches on stage_ready, and a PLANNING run with an accepted spec auto-opens its first decomposition loop), stage auto-retry (autoRetryLimit — default 2: the sweep retries a mechanically blocked stage up to this many times while requireHumanGates is false; 0 disables; decision blocks never auto-retry), and branch strategy. Distinct from updating name/repo/branch config. Pass expectedUpdatedAt for optimistic concurrency. Admin only.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "autoAdvanceStages": {
             "description": "Auto-launch the next launchable stage on harness.stage_ready (after decomposition and after each stage commit). Default true; set false to launch stages manually via request_harness_stage_run.",
@@ -14851,7 +14650,6 @@
       },
       "description": "Archive or restore a harness template (ADMIN; admin scope) — one setter over one field, so pass archived: true or false. Archiving removes it from the default registry listing and blocks edit, publish and launch; it is a SOFT delete, so the template, every version, and every run already launched from it are retained, and archived: false reverses it exactly. expectedUpdatedAt is REQUIRED. Setting the state it is already in is a no-op.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "archived": {
             "description": "true archives the template (it leaves the default registry listing and can no longer be edited, published or launched); false restores it. A soft delete either way — versions and launched runs are always retained.",
@@ -14887,7 +14685,6 @@
       },
       "description": "Change a member's role (admin scope; the op requires the OWNER role, and the last owner cannot be demoted). Governed by the membership pack (REQUIRE_APPROVAL by default). Concurrency is VALUE-CAS: pass `expectedRole` (the role you last read); `expectedUpdatedAt` is refused (M19.2). A value-CAS mismatch returns CONFLICT embedding the current role.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "expectedRole": {
             "description": "Value-CAS guard: the member's role you last read. If it changed since, the call fails with CONFLICT.",
@@ -14938,7 +14735,6 @@
       },
       "description": "Enable or disable a model profile (admin scope + ADMIN role). Converges to the target state (same action → same end state). Returns the updated profile.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "enabled": {
             "type": "boolean"
@@ -14971,7 +14767,6 @@
       },
       "description": "Enable or disable a provider connection (admin scope + ADMIN role). Converges to the target state. Returns the updated (redacted) connection.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "connectionId": {
             "minLength": 1,
@@ -15004,7 +14799,6 @@
       },
       "description": "Update a linked repo's sync config (admin scope + ADMIN role): board, default/closed columns, direction, and the label/assignee/comment sync toggles + enabled. Only provided fields change.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "boardId": {
             "minLength": 1,
@@ -15069,7 +14863,6 @@
       },
       "description": "Enable, pause, or resume the workspace's software-company automation (ADMIN). Enable and resume run the fail-closed preflight first — the whole harness suite must resolve every board/agent/release/policy binding and the managed-path gating eval must pass. That eval gate is currently blocked on AGE-188/235/236 (the only eval backend is the simulated in-process plane), so enable/resume are refused with the exact blockers until a real backend lands; fixture/simulator-only evidence can never enable the profile. Pause never runs preflight and never cancels active Agent Jobs, Harness Runs, deliveries, or releases — it only stops new work from starting.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "action": {
             "enum": [
@@ -15106,7 +14899,6 @@
       },
       "description": "Set the full list of assignees on a task. REPLACES existing assignees. All user IDs must be members of the task's workspace.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "taskId": {
             "description": "ID of the task.",
@@ -15140,7 +14932,6 @@
       },
       "description": "Set the full list of labels on a task. REPLACES existing labels; pass current+new to add, pass current-removed to delete. All label IDs must belong to the task's board.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "labelIds": {
             "description": "Full list of label IDs to apply. REPLACES all existing labels — pass current+new to add, pass current-removed to delete.",
@@ -15174,7 +14965,6 @@
       },
       "description": "Enable or disable a managed-execution worker (ADMIN — the R13 kill switch). Disabling sets disabledAt and prevents the worker from accepting NEW claims immediately (in-flight runs finish or are cancelled explicitly); re-enabling clears it. Each change writes an audit activity. Returns the updated worker.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "enabled": {
             "type": "boolean"
@@ -15207,7 +14997,6 @@
       },
       "description": "Start a shadow or canary experiment for an agent (admin scope + ADMIN role for a canary). The baseline is NEVER caller-supplied — it is the profile's currently pointed-to release. A CANARY carries the FULL M13.4 activation bar, checked atomically: the candidate's green gating EvalRun for its exact digest, policy pins equal to the current required revisions, its dataset version at the current regression head, an explicit admin gate (approvalRef), and candidate preflight against the eligible worker pool — each missing requirement is named, and a stale head/policy is refused (a canary is partial activation, never a way around activation). A RUNNING canary deterministically routes trafficFraction of eligible claims to the candidate WITHOUT moving the pointer; guardrails evaluated on the sweep cadence roll it back on a breach. SHADOW carries a lighter bar (effects are captured, zero real side effects) and routes no traffic. Returns the started experiment.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "agentProfileId": {
             "minLength": 1,
@@ -15307,7 +15096,6 @@
       },
       "description": "Start the authoring loop on a DRAFT harness run: moves it DRAFT → AUTHORING, seeds a run-owned draft Spec, and opens the authoring agent job bound to the run (the planner/reviewer run the review loop toward the run's spec). Requires an anchor task, a repository on the run, and active planner+reviewer agents (set via set_harness_agents). Approving the resulting plan-approval gate (UI / Control Tower — no MCP self-approval) accepts the run-owned Spec, pins its SpecVersion, and advances the run AUTHORING → PLANNING. Pass expectedUpdatedAt for optimistic concurrency. Returns the run's new status + the opened job id. Admin only.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "expectedUpdatedAt": {
             "description": "Optimistic concurrency guard: the entity's updatedAt from your last read. If it changed since, the call fails with CONFLICT and the error embeds the current state.",
@@ -15355,7 +15143,6 @@
       },
       "description": "Start the decomposition loop on a PLANNING harness run (its spec validated and pinned): opens the decomposition agent job bound to the run. The planner proposes ordered stages via submit_plan's harnessStages (each tracing the accepted spec criteria it delivers); the reviewer validates them. Approving the resulting plan-approval gate (UI / Control Tower — no MCP self-approval) checks that every accepted spec criterion is covered, then mints the ordered HarnessStages + 1:1 Work Orders, snapshots the registry, and advances the run PLANNING → READY (an uncovered plan is rejected, naming the orphaned criteria, and the gate stays pending). Requires an anchor task, a repository, and active planner+reviewer agents. The run stays PLANNING until the gate mints the stages. Pass expectedUpdatedAt for optimistic concurrency. Returns the run's status + the opened job id. Admin only.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "expectedUpdatedAt": {
             "description": "Optimistic concurrency guard: the entity's updatedAt from your last read. If it changed since, the call fails with CONFLICT and the error embeds the current state.",
@@ -15403,7 +15190,6 @@
       },
       "description": "Arm a one-time import of a linked repo's existing issues (admin scope + ADMIN role). The /api/cron/github driver advances it one page per tick. Idempotent — re-arming resets the cursor.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "repoLinkId": {
             "description": "Repo link to remove.",
@@ -15428,7 +15214,6 @@
       },
       "description": "Manually stop a RUNNING experiment (admin scope + ADMIN role). A CANARY stop is a rollback — it opens an incident and traffic falls back to the baseline (the pointer is never moved); a SHADOW stop simply ends capture. A non-RUNNING experiment is refused with CONFLICT. Use expectedRevision for a safe retry. Returns the terminated experiment.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "expectedRevision": {
             "maximum": 9007199254740991,
@@ -15461,7 +15246,6 @@
       },
       "description": "Executor: submit the execution result (summary + produced artifact ids) and hand off to the reviewer. Artifacts are validated against the workspace and linked to this run. Requires the claimed runId.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "artifactIds": {
             "description": "Artifact ids produced this run (validated + linked to the run).",
@@ -15549,7 +15333,6 @@
       },
       "description": "Attach a generated artifact (PRD candidate/synthesized, review report, stage plan, prompt template, execution summary, diff, or test evidence) to a harness run/stage — either capturing new redacted markdown (title+body) or classifying an existing visible Artifact (artifactId). Returns the link (null if the artifact store is unconfigured).",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "artifactId": {
             "anyOf": [
@@ -15864,7 +15647,6 @@
       },
       "description": "Reviewer: submit a verdict (APPROVED or CHANGES_REQUESTED) with structured findings on a harness stage_execution (or plan) job — the SAME M7 review contract, including severity-gated verdicts: every finding carries a REQUIRED severity 0-100 (>=60 blocks; below that a CHANGES_REQUESTED is recorded as APPROVED with follow-up findings). APPROVED opens the human acceptance gate; the findings normalize into the harness finding table so the validation gate + re-review continuity see them. Requires the claimed runId.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "expectedUpdatedAt": {
             "description": "Optimistic-concurrency guard: the job's updatedAt from your last read.",
@@ -16035,7 +15817,6 @@
       },
       "description": "Record a stage's structured execution output (changed files, commands run, tests run/not-run, risks, and the branch/commit it landed on) as a redacted execution-summary artifact + link the diff / test-evidence artifacts. Returns the post-capture validation gate (validatable + missing evidence + blocking-finding count) — the implicit validation request.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "branch": {
             "anyOf": [
@@ -16195,7 +15976,6 @@
       },
       "description": "Planner: submit (or revise) the plan. Freezes the plan as a new Work Order version and hands off to the reviewer. Requires the claimed runId. Pass resolutions to record how prior findings were fixed. For a harness decomposition job (templateKey 'decomposition'), pass harnessStages — the proposed ordered stages, each tracing the accepted spec criteria it delivers — which are persisted as the exact proposal the accept gate mints; harnessStages is rejected for any other workflow.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "acceptanceCriteria": {
             "description": "Required behavior, tests, and definition of done.",
@@ -16520,7 +16300,6 @@
       },
       "description": "Reviewer: submit a verdict (APPROVED or CHANGES_REQUESTED) with structured findings and the ids of any previously-addressed findings you now verify. Each finding carries a REQUIRED severity 0-100 (>=80 critical, 60-79 high, 40-59 medium, 20-39 low, <20 nit). Verdicts are severity-gated: CHANGES_REQUESTED stands only when a severity>=60 finding backs it (submitted now, or still open from a prior round) — otherwise it is recorded as APPROVED and the findings become follow-ups, so reject only for concrete blockers. On a HARNESS STAGE review every open blocking finding must be classified: re-raised in findings[], verified in verifiedFindingIds[], or deliberately left open in sustainedFindingIds[] — omitting one is rejected, because silence must never act as a veto. APPROVED opens a human approval gate. Requires the claimed runId.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "expectedUpdatedAt": {
             "description": "Optimistic-concurrency guard: the job's updatedAt from your last read.",
@@ -16691,7 +16470,6 @@
       },
       "description": "Mark a subtask complete or incomplete.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "completed": {
             "description": "New completion state.",
@@ -16721,7 +16499,6 @@
       },
       "description": "Move a harness stage to a new status through the GENERIC mover (admin scope + ADMIN role). Illegal transitions are refused verbatim by the ops core: COMMITTED (the commit boundary owns it), EXECUTING/VALIDATED (the execution loop owns them), moving a stage OUT of the execution family, and resuming a stage parked on an unaccepted blocking decision are all refused — this tool adds NO new legality. expectedUpdatedAt guards the write; a concurrent change returns CONFLICT embedding the current stage.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "expectedUpdatedAt": {
             "description": "Optimistic concurrency guard: the entity's updatedAt from your last read. If it changed since, the call fails with CONFLICT and the error embeds the current state.",
@@ -16787,7 +16564,6 @@
       },
       "description": "Restore one to 50 boards in a workspace atomically. Requires workspace admin access; already-active boards are idempotent no-ops.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "boardIds": {
             "description": "One to 50 unique board ids to restore. Discover archived ids via list_boards(workspaceId).",
@@ -16823,7 +16599,6 @@
       },
       "description": "Resume (unfreeze) a frozen harness run — restores its EXACT recorded prior status and clears the freeze. An operator/admin control (admin scope + admin role): an agent has no path to unfreeze its own run. Records the resuming actor + reason. Rejects a run that is not frozen. Admin only.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "expectedUpdatedAt": {
             "description": "Optimistic concurrency guard: the entity's updatedAt from your last read. If it changed since, the call fails with CONFLICT and the error embeds the current state.",
@@ -16871,7 +16646,6 @@
       },
       "description": "Unlink a repository from its board (admin scope). Synced tasks are kept; only the sync records are removed.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "repoLinkId": {
             "description": "Repo link to remove.",
@@ -16896,7 +16670,6 @@
       },
       "description": "Update an agent profile (admin scope + ADMIN role for a workspace agent). Only provided fields change. A trust-level RAISE on a workspace agent requires a passing gating eval for the agent's current configuration (policy-visible). expectedUpdatedAt guards the write; a stale value returns CONFLICT.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "agentId": {
             "description": "ID of the agent to update.",
@@ -17131,7 +16904,6 @@
       },
       "description": "Update the BugReport sidecar on a task whose board kind = BUGS. Only provided fields are changed; pass null to clear. Setting resolvedAt records BUG_RESOLVED; clearing it records BUG_REOPENED.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "actualBehavior": {
             "anyOf": [
@@ -17254,7 +17026,6 @@
       },
       "description": "Edit a comment's body. Only the comment's author can edit it.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "body": {
             "description": "New comment body (markdown, max 20k chars).",
@@ -17286,7 +17057,6 @@
       },
       "description": "Update fields on an existing contact. Only provided fields are changed; pass null to clear optional fields.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "company": {
             "anyOf": [
@@ -17373,7 +17143,6 @@
       },
       "description": "Update the Deal sidecar on a task whose board kind = CRM. Only provided fields are changed; pass null to clear. Creates the Deal row if it doesn't exist yet.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "amount": {
             "anyOf": [
@@ -17443,7 +17212,6 @@
       },
       "description": "Update a harness run's config/name (admin scope + ADMIN role): repo, baseBranch, allowedPaths, branch template/strategy, sequential, maxIterations, reviewerRequired. NOT a status mover — run lifecycle moves are owned by their loops. expectedUpdatedAt is the concurrency guard; a stale value returns CONFLICT embedding the current config.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "allowedPaths": {
             "description": "Path globs the run is allowed to touch (empty = unrestricted).",
@@ -17551,7 +17319,6 @@
       },
       "description": "Resolve a harness finding: mark it ACCEPTED / FIXED / REJECTED / VERIFIED / SUPERSEDED, or OPEN to reopen a fixed/verified finding on a regression (legal transitions only). REJECTED and SUPERSEDED require a resolutionNote. Pass expectedUpdatedAt for optimistic concurrency (stale → CONFLICT embedding the finding's current status/resolution). Reviewers reopening a regression should NOT set `adjudication` — it is for a human operator ruling on a parked finding, and it suppresses the reopen counter that drives the oscillation park. Setting `adjudication` raises this call to the ADMIN surface (admin token scope + workspace ADMIN role), like the other operator controls; a plain resolution stays at `write`.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "adjudication": {
             "description": "Set TRUE when a human operator is adjudicating this finding rather than the loop reopening it. A move back into OPEN normally counts as a reopen and feeds the oscillation park; an adjudication is the act that ENDS an oscillation, so it must not be counted as one. Defaults to false, so the reviewer's own reopen path is unchanged. Operator authority: setting it requires the workspace ADMIN role (and, over MCP, an admin-scoped token) — a plain resolution needs neither.",
@@ -17623,7 +17390,6 @@
       },
       "description": "Edit a harness template's DRAFT (ADMIN; admin scope). Published versions are immutable, so if the head is published this FORKS a new draft seeded from it and applies the edit there — the response's `forkedNewDraft` says which happened, and the new version number comes back with it. Send the whole definition document (it replaces the draft's), and/or the aggregate metadata (name/description/visibility). expectedUpdatedAt is REQUIRED: a stale value fails with CONFLICT carrying the template's current state, and a draft published out from under you fails with CONFLICT telling you to re-read and fork. Returns the resulting validation summary.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "definition": {
             "additionalProperties": false,
@@ -20234,7 +20000,6 @@
       },
       "description": "Update the Initiative sidecar on a task whose board kind = ROADMAP. Only provided fields are changed; pass null to clear.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "confidence": {
             "anyOf": [
@@ -20320,7 +20085,6 @@
       },
       "description": "Update fields on an existing project (name, status, owner, parent, goal statement, constraints, budget, success metrics). Only provided fields are changed; pass null to clear a nullable field. Needs the workspace ADMIN role. Pass expectedUpdatedAt (from your last read) for optimistic concurrency — a mismatch returns CONFLICT embedding the current state.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "budget": {
             "anyOf": [
@@ -20507,7 +20271,6 @@
       },
       "description": "Update fields on an existing task. Only provided fields are changed.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "description": {
             "description": "New markdown description.",
@@ -20573,7 +20336,6 @@
       },
       "description": "Update the Ticket sidecar on a task whose board kind = SUPPORT. Only provided fields are changed; pass null to clear.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "expectedUpdatedAt": {
             "description": "Optimistic concurrency guard: the entity's updatedAt from your last read. If it changed since, the call fails with CONFLICT and the error embeds the current entity.",
@@ -20675,7 +20437,6 @@
       },
       "description": "Rename a workspace and/or set its connected-session default repository (admin scope + ADMIN role). `defaultRepo` takes a normalized \"owner/name\" (lowercased at the write boundary; null clears it) — sessions fall back to it only when no project in the workspace carries an explicit REPO link. `slug` and `taskPrefix` are WRITE-ONCE — passing either is refused with INVALID_INPUT. Rename concurrency is VALUE-CAS, not timestamp CAS: pass `expectedName` (the name you last read); `expectedUpdatedAt` is refused (workspaces carry no updatedAt — M19.2). A value-CAS mismatch returns CONFLICT embedding the current state under error.current.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "defaultRepo": {
             "anyOf": [
@@ -20730,7 +20491,6 @@
       },
       "description": "Validate a harness template's draft (or a named version) WITHOUT writing anything — the SAME check that gates publishing, so `ok: true` here means publish_harness_template_version will not refuse on validation. Returns every finding with a stable code, severity (error blocks publish, warning never does), and a dotted path into the definition (e.g. steps.2.evidence). Secret-shaped literals are reported as an `secret_material` error: a template references credentials by alias only.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "harnessTemplateId": {
             "minLength": 1,
@@ -20758,7 +20518,6 @@
       },
       "description": "Verify a provider connection (ADMIN). Per-adapter: an api_key or custom endpoint is checked by a live server-side GET /v1/models probe (a model LISTING, never a model invocation) through the SSRF guard, and the raw signal is mapped to exactly one distinguished outcome — success, success-with-no-models, invalid credentials, unsupported endpoint, network failure, rate limited, model access denied, or provider temporarily unavailable — never a generic 'failed' or a false 'healthy'. A subscription connection has no server API to ping, so this recomputes its verification from the runner attestations instead. Returns the updated connection, the distinguished outcome (null for a subscription), and a redaction-safe human summary.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "connectionId": {
             "minLength": 1,
@@ -20787,7 +20546,6 @@
       },
       "description": "Freeze a Work Order's current draft into a new immutable version + Context Pack (with checksums). Returns the Work Order with its updated version list.",
       "inputSchema": {
-        "$schema": "http://json-schema.org/draft-07/schema#",
         "properties": {
           "idempotencyKey": {
             "maxLength": 200,