codex-orchestrator 2.0.2 → 2.0.4

package/internal-workflow/docs/agents/confidence-rubric.md

@@ -0,0 +1,65 @@
+# Confidence Rubric
+
+Use this rubric for coding skills that report findings, diagnose root causes, review specs, or decide whether an auto-fix is safe.
+
+Do not use fake numeric precision unless a skill has a concrete scoring reason. Prefer `high`, `medium`, and `low` with evidence.
+
+## High Confidence
+
+High confidence means the finding or diagnosis has direct evidence.
+
+Requires:
+
+- a concrete trigger path, code path, failing signal, or tool output
+- a clear explanation of why existing guards do not prevent the issue
+- no unresolved assumption that changes the conclusion
+
+Allowed actions:
+
+- report as a finding
+- block execution when the issue is safety-critical or spec-critical
+- auto-fix only when the fix is narrow, low-risk, local-patterned, and verifiable
+
+## Medium Confidence
+
+Medium confidence means the issue is likely but one explicit assumption remains.
+
+Requires:
+
+- strong local evidence
+- exactly what assumption remains
+- what evidence would promote or demote the finding
+
+Allowed actions:
+
+- report as a likely issue, risk, or execution concern
+- ask a targeted question when the unresolved assumption changes the fix
+- do not auto-fix unless new evidence raises confidence to high
+
+## Low Confidence
+
+Low confidence means the concern is plausible but not proven.
+
+Requires:
+
+- a clear label as uncertainty
+- the missing evidence or verification gap
+
+Allowed actions:
+
+- present as a question, risk, or verification gap
+- do not report as a proven bug
+- do not auto-fix
+
+## Auto-Fix Gate
+
+Auto-fix is allowed only when all are true:
+
+- confidence is high
+- root cause is clear
+- fix is narrow and low-risk
+- fix matches local project patterns
+- verification is available, or the edit is syntax-checkable and obviously safe
+- the change does not require a product decision
+
+If any condition is missing, report the issue with evidence and stop before editing.

package/internal-workflow/docs/agents/contract-test-ledger.md

@@ -0,0 +1,60 @@
+# Contract Test Ledger
+
+Use this shared ledger for behavior-changing work where a passing happy-path test could still miss a contract defect. The ledger turns review-class risks into testable obligations before implementation.
+
+## When Required
+
+Create or update a contract test ledger when the task changes any of these:
+
+- API, DTO, schema, serialization, persistence, or externally visible response shape
+- ordering, lifecycle events, state transitions, retries, idempotency, timeout, cancellation, or background jobs
+- cache keys, invalidation, state merge precedence, fallback behavior, profile/global/mobile overrides, defaults, or feature flags
+- evidence, trace, snapshot, audit, summary, aggregation, score, winner, or generated artifacts
+- shared behavior read by multiple callers, tenants, users, groups, children, or projections
+
+For narrow UI copy, docs-only, formatting, tests-only, or isolated styling changes, the ledger is not required.
+
+## Required Shape
+
+Keep the ledger compact. Use one row per invariant:
+
+```markdown
+## Contract Test Ledger
+
+| Invariant | Risk It Prevents | First Test / Proof | Status |
+| --- | --- | --- | --- |
+| <observable rule> | <real failure mode> | <exact test name/command or manual proof> | planned / red / green / blocked |
+```
+
+Rules:
+
+- The invariant must be observable through the public interface or the same seam real callers use.
+- The risk must name the concrete bug class, not a vague "edge case".
+- The first test/proof must fail before the fix unless the ledger records why a RED signal is impossible.
+- `blocked` requires the missing seam, fixture, service, or decision that prevents proof.
+- Keep the ledger current as implementation proceeds; do not backfill it only at the end.
+
+## Invariant Prompts
+
+Ask the relevant subset before the first RED test:
+
+- **Ordering:** What must happen before/after terminal events, snapshots, persistence writes, notifications, or cleanup?
+- **Precedence:** Which source wins among user input, profile, mobile, global, server, cache, AI, fallback, default, `null`, `false`, `0`, and empty objects?
+- **Threading:** Does each new field survive construction, normalization, cloning, retry, persistence reload, serialization, and every visible consumer?
+- **Runtime contract:** Do validation, internal types, persistence schema, API response, and consumers agree on names, units, nullability, enum values, and date/object formats?
+- **Retry/idempotency:** What changes on retry, and which snapshots, counters, streams, timestamps, writes, or side effects must be rebuilt instead of reused?
+- **Determinism:** When sort keys, timestamps, scores, priorities, or winners tie, what stable tie-breaker makes output repeatable?
+- **Evidence:** Which trace, audit, snapshot, Fresh-Context, summary, or generated artifact proves the behavior actually happened?
+- **Partial failure:** If a dependency times out, throws, returns stale data, or fails after a side effect, what durable state remains and who repairs it?
+- **Scope/cardinality:** Is data global, per-tenant, per-group, per-child, per-step, or per-item, and can one top-level field collapse multiple meaningful results?
+
+## Review Feedback Loop
+
+When code review finds a real contract defect, add or update one ledger row before fixing it:
+
+- `Invariant`: the rule the implementation violated
+- `Risk It Prevents`: the observed review finding
+- `First Test / Proof`: the regression test or proof that would have caught it
+- `Status`: `red` before the fix, then `green` after verification
+
+If no correct public seam exists for the regression test, record that as `blocked` and name the architecture/testability gap. Do not replace a missing seam with an implementation-detail test unless the task explicitly approves that tradeoff.

package/internal-workflow/docs/agents/review-gates.md

@@ -0,0 +1,42 @@
+# Review Gates
+
+This file owns review applicability. Review execution mechanics live in
+[`review-protocol.md`](review-protocol.md); approved-spec review shape lives in
+[`spec-implementer/references/review-loop.md`](../../skills/spec-implementer/references/review-loop.md).
+
+## Final Code Review
+
+Run final `$code-review` for behavior-changing work that affects:
+
+- medium/large shared business behavior;
+- API/DTO/schema, migration, persistence, auth, permission, payment, cache,
+  concurrency, background jobs, or shared-state contracts;
+- shared UI/navigation/middleware/core flows;
+- runtime logic across three or more files when it crosses an owner or
+  validation seam.
+
+Do not invoke it automatically for docs, copy, comments, tests-only or
+styling-only changes, formatting, renames, mechanical refactors, or isolated
+low-risk one-file fixes.
+
+`medium` is the normal review profile. API, persistence, statefulness, file
+count, or orchestration strengthens the review focus only when it creates an
+affected contract; none independently selects `high`.
+
+## Cleanup
+
+Cleanup is a lens inside the same final `$code-review`, never a separate gate.
+Use bounded cleanup by default. Amplify it only when the user, approved source,
+or repository policy names a concrete evidenced simplification risk; follow
+`../../skills/code-review/references/cleanup-lens.md` for that branch.
+
+After one consolidated repair, coordinator verification plus affected
+validation closes ordinary medium/low behavior-preserving findings. Use
+Closure only for the triggers in `review-protocol.md`.
+
+## Validation Depth
+
+For simple and medium work, run targeted behavior proof and the smallest
+affected integration check. Run a full repository suite only when explicitly
+required by repository policy, when broad contract fan-out cannot be isolated,
+or for a genuinely `high` task.

package/internal-workflow/docs/agents/review-protocol.md

@@ -0,0 +1,98 @@
+# Review Protocol
+
+This file owns mechanics shared by artifact and implementation review: bounded
+capsules, Full/Closure, defect lifecycle, no-progress, waiver, and the result
+envelope. Target Modules own authority, profile, topology, durable state, and
+outcome mapping.
+
+## Capsule
+
+Every reviewer receives only:
+
+- target kind/path and pinned revision;
+- one unanswered review question and assigned lenses;
+- authority, approved scope, and relevant evidence;
+- compact affected validation and current defect records;
+- for Closure, the repair diff and mapping from changed targets to defects and
+  affected contracts.
+
+Do not pass raw parent history, old revisions, repeated logs, or unrelated
+inventories. Reuse valid coverage for the same revision, question, and lenses.
+
+## Modes
+
+**Full** covers the complete assigned scope once. It is bounded to the settled
+target, changed owners, authority, and plausible affected contracts. It is not
+a repository-wide audit and does not activate unrelated lenses.
+
+**Closure** verifies a repair only when the finding is critical/high, affects a
+trust boundary, durable data, concurrency/idempotency, shared API/DTO/schema, or
+invalidates mandatory coverage. Closure stays with affected reviewer lineages
+and targets. Do not launch it merely because Full found an ordinary defect.
+
+A repair starts another Full only when mandatory-lens coverage became invalid.
+A live reviewer poll timeout is non-terminal and does not authorize duplicate
+review or cancellation; reconcile the recorded session first.
+
+## Defects
+
+Use one canonical record per distinct invariant and failure mechanism:
+
+```yaml
+id: REVIEW-CONC-003
+class: blocker | execution-risk | improvement
+status: open | fixed | verified | blocked | accepted-risk | superseded
+severity: critical | high | medium | low
+confidence: high | medium | low
+invariant: "<observable rule>"
+failure: "<concrete trigger and impact>"
+evidence: ["<target or source>"]
+repair: "<smallest sufficient change>"
+affected_targets: ["<path, section, contract, or lens>"]
+```
+
+Root assigns stable IDs and deduplicates only when both invariant and failure
+match. A superseded record must point to a distinct canonical replacement.
+Improvements never block. Only an execution risk may become `accepted-risk`,
+and only with explicit authority, reason, scope, and target revision. A blocker
+cannot be accepted or downgraded.
+
+Move blocking defects `open -> fixed -> verified`. For ordinary medium/low
+behavior-preserving repairs, root may verify after checking the cited failure
+path and affected validation. Closure-triggering defects require affected
+independent verification.
+
+In Closure, copy each supplied canonical defect's `id`, `class`, `invariant`,
+`failure`, and introduced target revision byte-for-byte. Never paraphrase those
+immutable fields while describing verification. Record the Closure decision in
+the status, status target revision, evidence, and repair-finding outcome fields.
+
+## Repair And Stop
+
+Repair compatible findings in one consolidated batch. Before repeating review,
+the target, evidence, repair, or source decision must change materially.
+Review count and elapsed time are audit signals, never approval or blocking
+conditions.
+
+Stop when repair requires a product/scope/owner decision, mandatory evidence or
+reviewer is unavailable, no substantive repair exists, or the same failure
+repeats without progress. Do not create micro-cycles for ordinary medium/low
+findings.
+
+Waive review only after explicit user instruction. Record skipped coverage and
+open defects. Waiver is not approval and accepts no defect automatically.
+
+## Result
+
+```text
+Review Mode: <Full | Closure>
+Mandatory Coverage: <covered lenses or gaps>
+Verified Defects: <IDs or None>
+Accepted Risks: <IDs, authority, reason or None>
+Open Defects: <IDs or None>
+```
+
+Target Modules add profile, authority, outcome, and checkpoint without
+redefining these fields. For a normal medium flow, do not report internal
+session accounting unless interruption, Closure, accepted risk, or another
+exception makes it relevant.

package/internal-workflow/docs/agents/tool-usage.md

@@ -0,0 +1,88 @@
+# Tool Usage
+
+## Flutter Debug Sessions
+
+Local Flutter UI work always uses a platform QA skill plus the runtime ownership gate:
+
+- Android emulator: use `$flutter-android-debug` as the lifecycle orchestrator and
+  `test-android-apps:android-emulator-qa` for navigation, interaction, UI trees,
+  screenshots, and logcat.
+- iOS Simulator: use `$flutter-ios-debug` for environment/login gates, navigation,
+  interaction, screenshots, and visual comparison.
+- Use `$flutter-attach-session` only to discover the runtime owner and perform
+  reload/restart when that session is safe for this agent to control.
+
+Before any install, launch, attach, reload, restart, terminate, or force-stop:
+
+1. Identify the project, target device, package/app, current PID, VM Service, runtime
+   owner, and expected backend/environment.
+2. Treat any live PID, VM Service, IDE debug adapter, `flutter run`, or visible app as
+   user-owned state.
+3. If an IDE/debug adapter or machine run owns DevFS, do not create a second attach
+   controller. Use the owning IDE/terminal reload action or ask the user to trigger it.
+   A standalone interactive terminal run is the only exception, and only when
+   discovery marks it attach-safe and the helper receives its confirmed PID.
+4. Use `r` for widget/layout/style/rendering changes and `R` for startup state,
+   dependency injection, providers, globals, routes, or initialization changes.
+5. For a safely attachable standalone runtime, pass the PID confirmed by
+   `discover --json` as `--expected-pid`; never execute the raw discovered attach command.
+6. Verify that the same PID and expected environment remain after each runtime or
+   navigation action.
+
+Requests to inspect, debug, navigate, capture screenshots, or verify UI do not authorize
+build/install, uninstall, app-data clearing, force-stop, process termination, replacement
+launch, or a new `flutter run`. Use those only when no live target exists and the user
+requested a fresh run, or after explicit approval to replace the current session.
+
+The platform QA skill owns navigation and evidence. Do not replace it with ad-hoc shell
+commands: use `test-android-apps:android-emulator-qa` for Android emulator UI trees,
+input, screenshots, and logcat, and `$flutter-ios-debug` for iOS Simulator environment
+gates, navigation, interaction, screenshots, and visual comparison.
+
+If the process disappears, stop and report it. Hot reload cannot restore a dead process,
+and launching the installed app may expose an older APK/IPA without the DevFS changes.
+
+For a cold launch, load the owning repository's launch configuration as the source of
+truth for flavor, dart-defines, package/app identity, and backend. This applies only
+after the cold-launch gate is satisfied.
+
+## External Research
+
+Prefer local evidence before external lookup. When external evidence is
+material to a coding decision, use the source that owns the claim:
+
+1. official documentation or specifications;
+2. first-party source code, changelogs, release notes, or issue trackers;
+3. first-party APIs or published schemas.
+
+Use secondary sources only to discover primary sources or identify a disputed
+interpretation. Record source version/date when freshness matters, separate
+sourced facts from inference, and say when current behavior cannot be confirmed.
+
+Keep one narrow lookup inline unless the user explicitly requests delegation or
+a durable artifact. Use `$research` for either explicit request, multi-source
+comparison, or material external contract synthesis. The skill owns the
+Research Capsule, named-agent route, claim-to-source artifact, root verification,
+and downstream handoff.
+
+## Context7
+
+Use Context7 only when the task genuinely requires precise, current documentation for a real package, framework, SDK, API, or implementation detail that cannot be confidently answered from local evidence or built-in knowledge.
+
+Prefer local inspection first:
+
+- source code
+- lockfiles and installed package versions
+- package manifests
+- existing tests and examples
+- local docs and configuration
+
+Do not use Context7 for:
+
+- routine coding
+- general language questions
+- simple refactors
+- repository-specific behavior
+- facts already visible in the project
+
+If Context7 is used, keep the lookup narrow and bring back only the details needed for the active task.

package/internal-workflow/evals/coding-skill-evals.json

@@ -0,0 +1,66 @@
+{
+  "schema_version": 1,
+  "purpose": "Cross-skill routing and handoff regressions only; skill-local behavior belongs in each owner's evals directory.",
+  "cases": [
+    {
+      "id": "route-feature-tdd",
+      "prompt": "Implement a behavior-changing feature in this repository.",
+      "expected": ["read local evidence", "activate tdd before behavior edits", "run affected validation"],
+      "forbidden": ["invent repository commands", "create planning artifacts without an execution gap"]
+    },
+    {
+      "id": "route-diagnosis-only",
+      "prompt": "Explain the root cause and do not fix it yet.",
+      "expected": ["use bug-root-cause-explainer", "cite concrete evidence", "stop before edits"],
+      "forbidden": ["edit files", "present an inference as proven"]
+    },
+    {
+      "id": "route-flaky-bug",
+      "prompt": "Debug a failure that only happens intermittently.",
+      "expected": ["use diagnosing-bugs", "improve reproduction or signal before repair"],
+      "forbidden": ["jump directly to a fix"]
+    },
+    {
+      "id": "profile-medium-default",
+      "prompt": "Implement one clear stateful feature touching an API, persistence, and several files.",
+      "expected": ["select medium", "implement directly when authority and proof are clear"],
+      "forbidden": ["select high from statefulness or file count", "manufacture a PRD, tickets, or spec"]
+    },
+    {
+      "id": "profile-high-threshold",
+      "prompt": "Classify one broad ordinary refactor and one change with irreversible data loss plus unclear recovery ownership.",
+      "expected": ["keep the broad ordinary refactor medium", "select high only for material consequence plus uncertainty"],
+      "forbidden": ["accumulate generic risk labels into high"]
+    },
+    {
+      "id": "planning-does-not-deliver",
+      "prompt": "Use wayfinder to resolve the decision frontier and publish the resulting tickets.",
+      "expected": ["stop after the planning package", "require separate delivery authorization"],
+      "forbidden": ["start implementation from publication or labels"]
+    },
+    {
+      "id": "spec-only-for-gap",
+      "prompt": "Choose the route for a clear implementation request and for the same request with an unresolved execution contract.",
+      "expected": ["route the clear request directly", "use implementation-spec-maker for the unresolved execution delta"],
+      "forbidden": ["require a spec for every medium change"]
+    },
+    {
+      "id": "ticket-direct-vs-graph",
+      "prompt": "Choose delivery for one deterministic approved ticket and for an approved dependency graph.",
+      "expected": ["allow direct execution for the single ticket", "use tickets-orchestrator for the graph"],
+      "forbidden": ["manufacture a wave-level spec"]
+    },
+    {
+      "id": "review-topology",
+      "prompt": "Choose reviewer topology for simple, medium, and high profiles.",
+      "expected": ["simple uses one reviewer_fast when gated", "medium uses one reviewer_standard", "high uses two disjoint reviewer_deep tracks"],
+      "forbidden": ["root self-review", "two reviewers for ordinary medium"]
+    },
+    {
+      "id": "review-closure-bounded",
+      "prompt": "A medium Full review reports one low cleanup issue and one high shared DTO contract defect.",
+      "expected": ["repair findings in one batch", "coordinator verifies the low repair", "Closure covers only the high affected contract"],
+      "forbidden": ["restart broad Full review", "launch a separate cleanup gate"]
+    }
+  ]
+}

package/internal-workflow/manifest.json

@@ -0,0 +1 @@
+{"evals":{"shared/coding-skill-evals":{"owner":null,"path":"evals/coding-skill-evals.json"},"skill/implementation-spec-review":{"owner":"implementation-spec-review","path":"skills/implementation-spec-review/evals/evals.json"},"skill/spec-implementer":{"owner":"spec-implementer","path":"skills/spec-implementer/evals/evals.json"}},"files":[{"mode":420,"path":"docs/agents/bug-workflow-routing.md","sha256":"a37c59676bcb8b938a7c82bce8a6ea5becf58ba367866d2712cd469391aea931","size":1603},{"mode":420,"path":"docs/agents/bugfix-quality-gate.md","sha256":"caaed6c923adfe56f4b6dd89222a83493ef4dd4e8bdd2bad694f8a11f93541ae","size":709},{"mode":420,"path":"docs/agents/coding-skill-routing.md","sha256":"958f4e7c7177c062b2a24fb7db2287be2cfa6e5281f2d156f7eb67b6cb3a0741","size":6434},{"mode":420,"path":"docs/agents/confidence-rubric.md","sha256":"42c947db2775380867e7adcfbdbe0f67b8f511b9ddf33a6eef8e1b0e995912c2","size":1883},{"mode":420,"path":"docs/agents/contract-test-ledger.md","sha256":"6f2327a40f218fbc746193c6440a69be08450938688b401f05163a3bb438a7f3","size":3779},{"mode":420,"path":"docs/agents/review-gates.md","sha256":"5ca48066b263cf869549c383814cfbdbbad711e5c247a8253bc2996d8fdec496","size":1857},{"mode":420,"path":"docs/agents/review-protocol.md","sha256":"9af5b44c545d76f3a048de424a4ccc78aa7e018bd193e4e5a787b2cef47af071","size":4086},{"mode":420,"path":"docs/agents/tool-usage.md","sha256":"b6ade11865a46a5a28d4823c1b70318453f4d350cfc62d2969aa06d73c47be4b","size":4321},{"mode":420,"path":"evals/coding-skill-evals.json","sha256":"22ba3bb4372cde747c3accc899f85dc67f190fd33113deb133392a1695f5ad99","size":3519},{"mode":420,"path":"operations/acceptance-proof/SKILL.md","sha256":"c33a04bf8dfcb59982f60b232633b0e48e9de4ec375cd02f52ec71f9fce30de6","size":440},{"mode":420,"path":"operations/ambiguity-review/SKILL.md","sha256":"20371f30015afef12a2b9dd608d21cc93a93f011873927f7a64e29b36a4fcf99","size":427},{"mode":420,"path":"operations/code-review/SKILL.md","sha256":"c415e4383dd7ddeb4b371a0a141a39c4296d95222b26ca3fcc51a34704e10f25","size":1246},{"mode":420,"path":"operations/implementation/SKILL.md","sha256":"6f0c9b900d252d9a833d7fdac6868d84900787debf2964e22ffbf86851af45d2","size":1461},{"mode":420,"path":"operations/spec-author/SKILL.md","sha256":"5170f275bd7bc346578028d74d5814a5edd36ccd90d0615c4d5e3d6b6f8af049","size":627},{"mode":420,"path":"operations/spec-review/SKILL.md","sha256":"6cb7b8ea245faa2587ebde07ebe3d364eff41d18d5464ccd2749cf2b9b49f701","size":653},{"mode":420,"path":"operations/triage/SKILL.md","sha256":"39e8301b90a59795dc2b93917d4674c22dc3e4380f011a54ade1de4c7cdc18ae","size":647},{"mode":420,"path":"profiles/analyst_deep.toml","sha256":"06335e3a13b07ef3d7deb9b546a6dad5d765edfa6dbf21c1c4d516e843a4352f","size":380},{"mode":420,"path":"profiles/implementer_standard.toml","sha256":"4074f45ea6fb615382de7ddf04e4ab824185e01932290b95764ff63a3711824e","size":750},{"mode":420,"path":"profiles/proof_agent.toml","sha256":"2fbaf1145a11cb5c574bf1ed7333187d284ed0c3c3d6260c628f82057b8b04e7","size":446},{"mode":420,"path":"profiles/reviewer_deep.toml","sha256":"15ef121b641265a85d0647c46f6dd93d4a9abd8c09bcc746c41e4ca6b4c9af41","size":648},{"mode":420,"path":"profiles/reviewer_standard.toml","sha256":"0a26b24f98b7e8d0ec049cb6fbe623d5da838b2fe71b5accd59bf364a9c97e5c","size":585},{"mode":420,"path":"schemas/ambiguity-review-v1.json","sha256":"48d946ad8e91bd1908d8993ef631b1701ad3cd7a79f34368a07942d7774afeff","size":524},{"mode":420,"path":"schemas/code-review-v1.json","sha256":"b11b266e0a7e0aa19eaf90ebf2b5c33f3bc7263e9c697cbf206e9865084e3439","size":2555},{"mode":420,"path":"schemas/implementation-report-v1.json","sha256":"a1b580dad03af9be74d895d38a2f6aa9398b6d772c944dc398dac5aca0630d2e","size":1573},{"mode":420,"path":"schemas/proof-report-v1.json","sha256":"1bf6c5b22b97ab3b659d961e21b0fc06869481405e1048ab9eadeeea212a8cf6","size":21949},{"mode":420,"path":"schemas/spec-author-v1.json","sha256":"49f945362d1184ad628584b91fbbe75323d13bafa4ca3876093387a9fca06214","size":420},{"mode":420,"path":"schemas/spec-review-v1.json","sha256":"fe9fdd389bcb4c3b7d19609184caf3af4b89e7c1d1e71a7ba0c55c72c5d5562d","size":1386},{"mode":420,"path":"schemas/triage-route-v1.json","sha256":"3ca7ed29237f42e12567797d145d90dd4f1f48fa1e81a7da67434c19747753d8","size":5894},{"mode":420,"path":"skills/acceptance-proof/SKILL.md","sha256":"5a0f2dcd62a43da86a7627c70c86bde86e78c06929675073ee37727e3f06fc65","size":1683},{"mode":420,"path":"skills/acceptance-proof/agents/openai.yaml","sha256":"d602d9f2e1bf618a71171729c6f354dd137c939cb433bd49afec678360e1189f","size":265},{"mode":420,"path":"skills/acceptance-proof/references/android.md","sha256":"b9396d1327ffc19f91b73c2470222871e403ea1d0f3de58025e689a92691a842","size":3022},{"mode":420,"path":"skills/acceptance-proof/references/browser.md","sha256":"ceefa4fd7db475b6162368511c2742b6d9c6af58d48e7d5d8a773ed5cc3938c4","size":2281},{"mode":420,"path":"skills/acceptance-proof/references/ios.md","sha256":"ae18be2632e1f7c3aa0810a1bafafc7760c807b30b1269408825aa14c224d493","size":2486},{"mode":420,"path":"skills/acceptance-proof/tools/android-lease.mjs","sha256":"982c426dd5b3a90f74e5120783b3a24fa98d8218d0f4140e756187865b48d708","size":11393},{"mode":420,"path":"skills/acceptance-proof/tools/ios-lease.mjs","sha256":"6e1e0d95c6a8b2d42c34de05bd4446eac23e3916bffccb027f236b8a8a1809a3","size":13834},{"mode":420,"path":"skills/agent-auto/SKILL.md","sha256":"450c28f7a712f881cdf9f3e48555a47022b46f79df875bac35843a1c15135451","size":1415},{"mode":420,"path":"skills/agent-auto/agents/openai.yaml","sha256":"79a70421300891e3130fc7535c4aa37e01931ff353fbd7b83262c6fcf2032b71","size":266},{"mode":420,"path":"skills/code-debugger/SKILL.md","sha256":"56a403b2cd9a3dad4b48d06ab05a9b99fde97261b7380c2cd9ffd82b2c3cbb53","size":7725},{"mode":420,"path":"skills/code-debugger/agents/openai.yaml","sha256":"8dd2f301bee632585371bd6f62dbdcdec95201f553392515342da5d0fc563305","size":322},{"mode":420,"path":"skills/code-review/SKILL.md","sha256":"5cd395731f598d5c88319f74ea4fe83ae7678fa1bd42cd45acc6246b67467289","size":15805},{"mode":420,"path":"skills/code-review/agents/openai.yaml","sha256":"c2697212427a5e119d9127e2f9000594e6c2eb7696e7a40604da7149c79ce478","size":278},{"mode":420,"path":"skills/code-review/references/bug-classes.md","sha256":"1f8648af9914cbd7553d045915f959df0f3b50a88e97c3beefeb955342e7b12d","size":4062},{"mode":420,"path":"skills/code-review/references/cleanup-lens.md","sha256":"6406350bf8ff00f9d2de2d5a8871aa1a9ab0efa33dcc35453eeebe9c96623e69","size":2846},{"mode":420,"path":"skills/code-review/references/framework-lenses.md","sha256":"ca9e7cb09f32f729cec5522e8ff7c7dcc43f76ef986e701e9c0a226514912cd8","size":1820},{"mode":420,"path":"skills/code-review/references/targeted-recipes.md","sha256":"921b422958c97e606637d3fa2d2628239d9176f0316c7e45bbc55b20a22fe3c1","size":2564},{"mode":420,"path":"skills/diagnosing-bugs/SKILL.md","sha256":"9a3457d4f12e3810041def93456a0c020df0842200737bf7ecaf06471991c16c","size":9097},{"mode":420,"path":"skills/diagnosing-bugs/agents/openai.yaml","sha256":"eca84840bc193ce63cc7aad93d9e7b5f2541739b7bf682e5fa9eded3a8060787","size":262},{"mode":420,"path":"skills/diagnosing-bugs/scripts/hitl-loop.template.sh","sha256":"b2932630950e5210075bcd6f850e5accf30c101c5367b29eac3a29b4dd8084c8","size":1164},{"mode":420,"path":"skills/implementation-spec-maker/SKILL.md","sha256":"035cb829574c92c8342ae0a4a6f04866802f3724b6ba16583ae8c5724d6d45f9","size":7751},{"mode":420,"path":"skills/implementation-spec-maker/agents/openai.yaml","sha256":"a4457f3e2f08cb07694855d362103b6a628c82b262950409268145348e2d91dd","size":363},{"mode":420,"path":"skills/implementation-spec-maker/references/source-modes.md","sha256":"471f0f58668b414438effbf91398022327fcbd43f40cc0801080994fe02bda3b","size":1920},{"mode":420,"path":"skills/implementation-spec-maker/references/spec-template.md","sha256":"0ba380125eb2e0c114aed6b0f064ac2643edf75bb78c10de72b741776edee6ff","size":5319},{"mode":420,"path":"skills/implementation-spec-review/SKILL.md","sha256":"d4c32638f0bf93766dda1e9d5eaa072c279d4658d421482da78ccc0cfd2e13c7","size":5270},{"mode":420,"path":"skills/implementation-spec-review/agents/openai.yaml","sha256":"600f9cc4e4f42596e3bf48601a508ddf055efcc396478d5e339d024973f4fb05","size":277},{"mode":420,"path":"skills/implementation-spec-review/evals/evals.json","sha256":"d0f73eba5cb0ce34f50f43f80094a3f0cf3aa1caa1b5eb859906721159aee575","size":1114},{"mode":420,"path":"skills/implementation-spec-review/references/review-loop.md","sha256":"9d9e9c233ba08e256b158a7fe740ffe5ffd1893960c2076254e44e0e49083630","size":3901},{"mode":420,"path":"skills/small-task-implementer/SKILL.md","sha256":"6b81bf9d85b2f6c8253099cfe766f67cad312e3eb38b014ed2bedb007df467fc","size":4279},{"mode":420,"path":"skills/small-task-implementer/agents/openai.yaml","sha256":"81569b6dfd97de60b53f092528140a5e5043f9adcc9262debff7bc912e278958","size":261},{"mode":420,"path":"skills/spec-implementer/SKILL.md","sha256":"70f65edddebc788a21dbb5bc6cb8f60a297e0364c8e4754f3ba80cbf1ba210a0","size":5538},{"mode":420,"path":"skills/spec-implementer/agents/openai.yaml","sha256":"84ef664ae3538e264fe6746917f081d34eac0738dc28764ddf27d7a448b3615d","size":341},{"mode":420,"path":"skills/spec-implementer/evals/evals.json","sha256":"23d73ddd6e2205b601a36cd07a7dd5ee228ae587d5b9cc5adab9b0993a312ba8","size":1361},{"mode":420,"path":"skills/spec-implementer/references/review-loop.md","sha256":"5161c4c586144cdf0aba828c459c57b9e8dc404a06990b3a887b51648fe1fcb9","size":4311},{"mode":420,"path":"skills/tdd/SKILL.md","sha256":"9e046610c341be0c770d4196f98c95cdb673d477bbbbcdfcca499dd5da6071d0","size":4146},{"mode":420,"path":"skills/tdd/agents/openai.yaml","sha256":"cc49a11a2c08733862d1a406123cda7a050d4a70fa92a4b3ec37f318694b1581","size":301},{"mode":420,"path":"skills/tdd/interface-design.md","sha256":"764c5ff0e3fa6b4ab7095eb65ccc7201e090baf19fa16051dfdb72c06d27417d","size":653},{"mode":420,"path":"skills/tdd/mocking.md","sha256":"3ceb807fdf4a47d6a93d4d9a891e5ba6d362a6247bd08adc451feebfc17361ef","size":1481},{"mode":420,"path":"skills/tdd/refactoring.md","sha256":"54fced22dd1911b7094c3fe7979b7c1a40d40be307482c4adf7dc0588f27d6cc","size":387},{"mode":420,"path":"skills/tdd/tests.md","sha256":"6773173a074569b2e51653bd7b95c097d602dbb4815d1cf1c87344705c4e0d1c","size":2228},{"mode":420,"path":"skills/triage/AGENT-BRIEF.md","sha256":"053cd013e1c2c9275111aa6e4b5a12e9838f8d6cd888e8ca0ccaaac576fd74df","size":7070},{"mode":420,"path":"skills/triage/OUT-OF-SCOPE.md","sha256":"8ed8cf27833444060c81b3961a83c0e3d8e6cf2fcb2ddf6f8b07c6655cbb0d85","size":4282},{"mode":420,"path":"skills/triage/SKILL.md","sha256":"5c7c84189fd5146ec1ae55a5669c74372ed7bce588fa7b8523417aa55b312a2e","size":8273},{"mode":420,"path":"skills/triage/agents/openai.yaml","sha256":"466bc430f95132bc6b28d077d50865d4b1fd9218307582343c35b0baf66d7894","size":283}],"generationHash":"a66ee20f05adca0ffc042d75e0121bf0b8a8679c67d6eecb933acd8f2af4ca1d","operations":{"acceptance-proof":{"dependencySkills":[],"entry":"operations/acceptance-proof/SKILL.md","files":["operations/acceptance-proof/SKILL.md","profiles/proof_agent.toml","schemas/proof-report-v1.json","skills/acceptance-proof/SKILL.md","skills/acceptance-proof/agents/openai.yaml","skills/acceptance-proof/references/android.md","skills/acceptance-proof/references/browser.md","skills/acceptance-proof/references/ios.md","skills/acceptance-proof/tools/android-lease.mjs","skills/acceptance-proof/tools/ios-lease.mjs"],"id":"acceptance-proof","outputSchema":"schemas/proof-report-v1.json","policy":{"approvalCeiling":"never","cwdClass":"worktree","externalWrite":false,"mcpTools":[],"network":"deny","networkHosts":[],"runnerPostcondition":"proof-only","sandboxMode":"workspace-write","worktreeAccess":"write","writableRootClasses":["worktree"]},"profile":"proof_agent","resources":[],"sourceSkill":"acceptance-proof"},"ambiguity-review":{"dependencySkills":[],"entry":"operations/ambiguity-review/SKILL.md","files":["docs/agents/confidence-rubric.md","operations/ambiguity-review/SKILL.md","profiles/reviewer_deep.toml","schemas/ambiguity-review-v1.json"],"id":"ambiguity-review","outputSchema":"schemas/ambiguity-review-v1.json","policy":{"approvalCeiling":"never","cwdClass":"worktree","externalWrite":false,"mcpTools":[],"network":"deny","networkHosts":[],"runnerPostcondition":"report-only","sandboxMode":"read-only","worktreeAccess":"read-only","writableRootClasses":[]},"profile":"reviewer_deep","resources":["docs/agents/confidence-rubric.md"],"sourceSkill":null},"code-review":{"dependencySkills":[],"entry":"operations/code-review/SKILL.md","files":["docs/agents/confidence-rubric.md","docs/agents/contract-test-ledger.md","docs/agents/review-gates.md","docs/agents/review-protocol.md","operations/code-review/SKILL.md","profiles/reviewer_standard.toml","schemas/code-review-v1.json","skills/code-review/SKILL.md","skills/code-review/agents/openai.yaml","skills/code-review/references/bug-classes.md","skills/code-review/references/cleanup-lens.md","skills/code-review/references/framework-lenses.md","skills/code-review/references/targeted-recipes.md"],"id":"code-review","outputSchema":"schemas/code-review-v1.json","policy":{"approvalCeiling":"never","cwdClass":"worktree","externalWrite":false,"mcpTools":[],"network":"deny","networkHosts":[],"runnerPostcondition":"report-only","sandboxMode":"read-only","worktreeAccess":"read-only","writableRootClasses":[]},"profile":"reviewer_standard","resources":["docs/agents/confidence-rubric.md","docs/agents/contract-test-ledger.md","docs/agents/review-gates.md","docs/agents/review-protocol.md"],"sourceSkill":"code-review"},"implementation":{"dependencySkills":["code-debugger","diagnosing-bugs","small-task-implementer","tdd"],"entry":"operations/implementation/SKILL.md","files":["docs/agents/bug-workflow-routing.md","docs/agents/coding-skill-routing.md","docs/agents/contract-test-ledger.md","docs/agents/review-gates.md","docs/agents/tool-usage.md","operations/implementation/SKILL.md","profiles/implementer_standard.toml","schemas/implementation-report-v1.json","skills/agent-auto/SKILL.md","skills/agent-auto/agents/openai.yaml","skills/code-debugger/SKILL.md","skills/code-debugger/agents/openai.yaml","skills/diagnosing-bugs/SKILL.md","skills/diagnosing-bugs/agents/openai.yaml","skills/diagnosing-bugs/scripts/hitl-loop.template.sh","skills/small-task-implementer/SKILL.md","skills/small-task-implementer/agents/openai.yaml","skills/tdd/SKILL.md","skills/tdd/agents/openai.yaml","skills/tdd/interface-design.md","skills/tdd/mocking.md","skills/tdd/refactoring.md","skills/tdd/tests.md"],"id":"implementation","outputSchema":"schemas/implementation-report-v1.json","policy":{"approvalCeiling":"never","cwdClass":"worktree","externalWrite":false,"mcpTools":[],"network":"deny","networkHosts":[],"runnerPostcondition":"change-set","sandboxMode":"workspace-write","worktreeAccess":"write","writableRootClasses":["worktree"]},"profile":"implementer_standard","resources":["docs/agents/bug-workflow-routing.md","docs/agents/coding-skill-routing.md","docs/agents/contract-test-ledger.md","docs/agents/review-gates.md","docs/agents/tool-usage.md"],"sourceSkill":"agent-auto"},"spec-author":{"dependencySkills":[],"entry":"operations/spec-author/SKILL.md","files":["docs/agents/confidence-rubric.md","docs/agents/contract-test-ledger.md","operations/spec-author/SKILL.md","profiles/implementer_standard.toml","schemas/spec-author-v1.json","skills/implementation-spec-maker/SKILL.md","skills/implementation-spec-maker/agents/openai.yaml","skills/implementation-spec-maker/references/source-modes.md","skills/implementation-spec-maker/references/spec-template.md"],"id":"spec-author","outputSchema":"schemas/spec-author-v1.json","policy":{"approvalCeiling":"never","cwdClass":"target-state","externalWrite":false,"mcpTools":[],"network":"deny","networkHosts":[],"runnerPostcondition":"spec-only","sandboxMode":"workspace-write","worktreeAccess":"write","writableRootClasses":["target-state"]},"profile":"implementer_standard","resources":["docs/agents/confidence-rubric.md","docs/agents/contract-test-ledger.md"],"sourceSkill":"implementation-spec-maker"},"spec-review":{"dependencySkills":[],"entry":"operations/spec-review/SKILL.md","files":["docs/agents/confidence-rubric.md","docs/agents/contract-test-ledger.md","docs/agents/review-protocol.md","operations/spec-review/SKILL.md","profiles/reviewer_deep.toml","schemas/spec-review-v1.json","skills/implementation-spec-review/SKILL.md","skills/implementation-spec-review/agents/openai.yaml","skills/implementation-spec-review/references/review-loop.md"],"id":"spec-review","outputSchema":"schemas/spec-review-v1.json","policy":{"approvalCeiling":"never","cwdClass":"worktree","externalWrite":false,"mcpTools":[],"network":"deny","networkHosts":[],"runnerPostcondition":"report-only","sandboxMode":"read-only","worktreeAccess":"read-only","writableRootClasses":[]},"profile":"reviewer_deep","resources":["docs/agents/confidence-rubric.md","docs/agents/contract-test-ledger.md","docs/agents/review-protocol.md"],"sourceSkill":"implementation-spec-review"},"triage":{"dependencySkills":[],"entry":"operations/triage/SKILL.md","files":["docs/agents/coding-skill-routing.md","operations/triage/SKILL.md","profiles/analyst_deep.toml","schemas/triage-route-v1.json","skills/triage/AGENT-BRIEF.md","skills/triage/OUT-OF-SCOPE.md","skills/triage/SKILL.md","skills/triage/agents/openai.yaml"],"id":"triage","outputSchema":"schemas/triage-route-v1.json","policy":{"approvalCeiling":"never","cwdClass":"worktree","externalWrite":false,"mcpTools":[],"network":"deny","networkHosts":[],"runnerPostcondition":"report-only","sandboxMode":"read-only","worktreeAccess":"read-only","writableRootClasses":[]},"profile":"analyst_deep","resources":["docs/agents/coding-skill-routing.md"],"sourceSkill":"triage"}},"profiles":{"analyst_deep":"profiles/analyst_deep.toml","implementer_standard":"profiles/implementer_standard.toml","proof_agent":"profiles/proof_agent.toml","reviewer_deep":"profiles/reviewer_deep.toml","reviewer_standard":"profiles/reviewer_standard.toml"},"skills":{"acceptance-proof":{"entry":"skills/acceptance-proof/SKILL.md","files":["skills/acceptance-proof/SKILL.md","skills/acceptance-proof/agents/openai.yaml","skills/acceptance-proof/references/android.md","skills/acceptance-proof/references/browser.md","skills/acceptance-proof/references/ios.md","skills/acceptance-proof/tools/android-lease.mjs","skills/acceptance-proof/tools/ios-lease.mjs"],"metadata":"skills/acceptance-proof/agents/openai.yaml"},"agent-auto":{"entry":"skills/agent-auto/SKILL.md","files":["skills/agent-auto/SKILL.md","skills/agent-auto/agents/openai.yaml"],"metadata":"skills/agent-auto/agents/openai.yaml"},"code-debugger":{"entry":"skills/code-debugger/SKILL.md","files":["skills/code-debugger/SKILL.md","skills/code-debugger/agents/openai.yaml"],"metadata":"skills/code-debugger/agents/openai.yaml"},"code-review":{"entry":"skills/code-review/SKILL.md","files":["skills/code-review/SKILL.md","skills/code-review/agents/openai.yaml","skills/code-review/references/bug-classes.md","skills/code-review/references/cleanup-lens.md","skills/code-review/references/framework-lenses.md","skills/code-review/references/targeted-recipes.md"],"metadata":"skills/code-review/agents/openai.yaml"},"diagnosing-bugs":{"entry":"skills/diagnosing-bugs/SKILL.md","files":["skills/diagnosing-bugs/SKILL.md","skills/diagnosing-bugs/agents/openai.yaml","skills/diagnosing-bugs/scripts/hitl-loop.template.sh"],"metadata":"skills/diagnosing-bugs/agents/openai.yaml"},"implementation-spec-maker":{"entry":"skills/implementation-spec-maker/SKILL.md","files":["skills/implementation-spec-maker/SKILL.md","skills/implementation-spec-maker/agents/openai.yaml","skills/implementation-spec-maker/references/source-modes.md","skills/implementation-spec-maker/references/spec-template.md"],"metadata":"skills/implementation-spec-maker/agents/openai.yaml"},"implementation-spec-review":{"entry":"skills/implementation-spec-review/SKILL.md","files":["skills/implementation-spec-review/SKILL.md","skills/implementation-spec-review/agents/openai.yaml","skills/implementation-spec-review/references/review-loop.md"],"metadata":"skills/implementation-spec-review/agents/openai.yaml"},"small-task-implementer":{"entry":"skills/small-task-implementer/SKILL.md","files":["skills/small-task-implementer/SKILL.md","skills/small-task-implementer/agents/openai.yaml"],"metadata":"skills/small-task-implementer/agents/openai.yaml"},"spec-implementer":{"entry":"skills/spec-implementer/SKILL.md","files":["skills/spec-implementer/SKILL.md","skills/spec-implementer/agents/openai.yaml","skills/spec-implementer/references/review-loop.md"],"metadata":"skills/spec-implementer/agents/openai.yaml"},"tdd":{"entry":"skills/tdd/SKILL.md","files":["skills/tdd/SKILL.md","skills/tdd/agents/openai.yaml","skills/tdd/interface-design.md","skills/tdd/mocking.md","skills/tdd/refactoring.md","skills/tdd/tests.md"],"metadata":"skills/tdd/agents/openai.yaml"},"triage":{"entry":"skills/triage/SKILL.md","files":["skills/triage/AGENT-BRIEF.md","skills/triage/OUT-OF-SCOPE.md","skills/triage/SKILL.md","skills/triage/agents/openai.yaml"],"metadata":"skills/triage/agents/openai.yaml"}},"sourceFingerprint":"a8849a4c47b3adcb10239694bc0404eb89df4dbf8ae1e6ca7be689c141c5e316","version":2}

package/internal-workflow/operations/acceptance-proof/SKILL.md

@@ -0,0 +1,9 @@
+# Acceptance Proof Operation
+
+Follow the packaged [Acceptance Proof skill](../../skills/acceptance-proof/SKILL.md).
+Never change product behavior or external state. Return only
+`schemas/proof-report-v1.json`.
+
+The Runner already supplied the exact schema through `--output-schema`. Do not
+search for, open, or infer a repository-relative schema file; inspect only the
+frozen criteria, changed targets, checks, and requested proof evidence.

package/internal-workflow/operations/ambiguity-review/SKILL.md

@@ -0,0 +1,5 @@
+# Fresh Ambiguity Review Operation
+
+Independently verify that a supplied waiting candidate contains at least two materially different product outcomes and no source-authorized choice. Technical, architecture, test, and tool choices are never product ambiguity. Do not edit files or external state. Return only `schemas/ambiguity-review-v1.json`.
+
+Apply the packaged [confidence rubric](../../docs/agents/confidence-rubric.md).

package/internal-workflow/operations/code-review/SKILL.md

@@ -0,0 +1,23 @@
+# Code Review Operation
+
+You are already the independent reviewer selected by the Runner. Follow the
+packaged [Code Review skill](../../skills/code-review/SKILL.md) inline with
+correctness and spec/standards lenses. Cleanup is its bounded lens, never a
+separate operation. Use the declared resources for
+[confidence](../../docs/agents/confidence-rubric.md),
+[contract tests](../../docs/agents/contract-test-ledger.md),
+[review applicability](../../docs/agents/review-gates.md), and
+[Full/Closure mechanics](../../docs/agents/review-protocol.md). The supplied
+review capsule is the exact target and authority.
+
+For Closure, copy every supplied canonical defect's `id`, `class`, `invariant`,
+`failure`, and `introducedTargetRevision` exactly. Do not paraphrase immutable
+defect fields; express verification through the allowed status, revision,
+evidence, and repair-finding outcome fields.
+
+Use `needs-work` for concrete defects that the bounded implementation cycle can
+repair. Reserve `rejected` for a target or authority that cannot safely proceed
+through the normal repair lifecycle.
+
+Do not launch another reviewer, edit files, repair findings, or mutate external
+state. Return only `schemas/code-review-v1.json` with operation `code-review`.

package/internal-workflow/operations/implementation/SKILL.md

@@ -0,0 +1,24 @@
+# Implementation Operation
+
+Follow [Agent Auto](../../skills/agent-auto/SKILL.md) as the Runner adapter.
+Use the packaged [coding routing](../../docs/agents/coding-skill-routing.md)
+only for direct implementation, TDD, bug-routing, evidence, and affected
+validation. Read [TDD](../../skills/tdd/SKILL.md) before behavior changes. For
+a confirmed bug use [Code Debugger](../../skills/code-debugger/SKILL.md); use
+[Diagnosing Bugs](../../skills/diagnosing-bugs/SKILL.md) only when a reliable
+failing signal is missing. A tiny task may use
+[Small Task Implementer](../../skills/small-task-implementer/SKILL.md) only
+after its Fit Gate. Apply the declared
+[bug routing](../../docs/agents/bug-workflow-routing.md),
+[contract ledger](../../docs/agents/contract-test-ledger.md),
+[review gate](../../docs/agents/review-gates.md), and
+[tool policy](../../docs/agents/tool-usage.md) only when their branch is active.
+
+The issue is already authorized for implementation. Do not start planning,
+ticket publication, implementation-spec authoring, independent review, or
+delivery. The Runner owns review, checks, commits, publication, retries, and
+external state. Never commit, push, publish, mutate GitHub, or expose
+credentials. In the final report, `changedFiles` is the complete current product
+change set across all implementation cycles, not only files touched in this
+attempt; exclude Runner-owned proof artifacts. Return only
+`schemas/implementation-report-v1.json`.

package/internal-workflow/operations/spec-author/SKILL.md

@@ -0,0 +1,12 @@
+# Spec Author Operation
+
+Follow the authoring and minimum-solution contract in the packaged
+[Implementation Spec Maker](../../skills/implementation-spec-maker/SKILL.md)
+for the supplied issue authority. Use the declared
+[confidence rubric](../../docs/agents/confidence-rubric.md) and
+[contract ledger](../../docs/agents/contract-test-ledger.md) only when
+applicable.
+The Runner owns artifact review, revision state, and approval, so do not invoke
+the skill's review/save workflow or create reviewer state. Write the complete
+revision only to the Runner-provided spec artifact location and return
+`schemas/spec-author-v1.json`.

package/internal-workflow/operations/spec-review/SKILL.md

@@ -0,0 +1,12 @@
+# Spec Review Operation
+
+You are already the independent reviewer selected and persisted by the Runner.
+Follow the packaged
+[Implementation Spec Review](../../skills/implementation-spec-review/SKILL.md)
+inline and use its owner-local review loop only for semantics matching the
+supplied mode and immutable state. Apply the declared
+[confidence rubric](../../docs/agents/confidence-rubric.md),
+[contract ledger](../../docs/agents/contract-test-ledger.md), and
+[review protocol](../../docs/agents/review-protocol.md). Do not launch another
+reviewer, edit the spec, change review state, or mutate external state. Return
+only `schemas/spec-review-v1.json`.

package/internal-workflow/operations/triage/SKILL.md

@@ -0,0 +1,12 @@
+# Triage Operation
+
+Inspect the supplied issue and repository evidence without edits or external
+writes. Use the packaged
+[coding routing](../../docs/agents/coding-skill-routing.md) for the distinction
+between a direct deterministic implementation and a real execution gap. Use
+the packaged [Triage skill](../../skills/triage/SKILL.md) only for evidence
+discipline; its labels and tracker mutations are outside this operation.
+
+Return exactly one package route in `schemas/triage-route-v1.json`: direct,
+spec-required, awaiting-user for material product ambiguity, or a typed
+blocker. Technical implementation choices never require awaiting-user.

package/internal-workflow/profiles/analyst_deep.toml

@@ -0,0 +1,9 @@
+name = "analyst_deep"
+description = "Deep read-only analysis for ambiguous architecture, contracts, and root causes."
+nickname_candidates = ["Sage"]
+model = "gpt-5.6-sol"
+model_reasoning_effort = "xhigh"
+sandbox_mode = "read-only"
+developer_instructions = """
+Analyze only. Build evidence-backed conclusions and recommendations; leave final decisions and edits to the parent.
+"""

package/internal-workflow/profiles/implementer_standard.toml

@@ -0,0 +1,9 @@
+name = "implementer_standard"
+description = "Write-capable implementation worker for one approved, bounded ticket slice."
+nickname_candidates = ["Forge", "Mason", "Builder"]
+model = "gpt-5.6-sol"
+model_reasoning_effort = "medium"
+sandbox_mode = "workspace-write"
+developer_instructions = """
+Implement only the assigned approved ticket slice through its observable interface. Respect the exact write scope, tests, exclusions, and stop conditions. You are not alone in the repository: preserve unrelated and concurrent changes, never revert work you do not own, and report overlap before editing. Use behavior-first proof for behavior changes and return changed files, acceptance proof, skipped checks, risks, and blockers to the root integrator.
+"""

package/internal-workflow/profiles/proof_agent.toml

@@ -0,0 +1,8 @@
+name = "proof_agent"
+description = "Write-capable contained proof worker; Runner enforces proof-only postconditions."
+model = "gpt-5.6-sol"
+model_reasoning_effort = "high"
+sandbox_mode = "workspace-write"
+developer_instructions = """
+Prove only the frozen criteria. Write only proof evidence requested by the Runner, never product behavior, Git history, GitHub state, publication state, or credentials. Return the exact supplied JSON schema.
+"""