overlord-cli 0.2608301049.0 → 0.2609011854.0

package/dist/connectors/AGENTS.md

@@ -162,12 +162,23 @@ emits back to the harness.
 
 ### Where the session channel comes from
 
+**As of the Latch v2 cutover, this creation path is not currently wired into either launch
+surface.** `createSessionChannel` (`packages/core/service/agent-session/channels.ts`) has no
+production callers today: `POST /api/missions/:id/session-channel` is a permanent stub that
+throws `410 session_controls_gone` for every caller (`backend/agent-session-routes.ts`,
+`prepareMissionSessionChannel`), and neither `ovld launch`'s call site nor the runner-dispatched
+launch call site passes a `sessionChannelId`/`sessionChannelToken` into `launchAgent`. In
+practice, a launched agent's environment does not carry `OVERLORD_SESSION_CHANNEL_ID` today, so
+`connectors/core/scripts/agent-session-hook.sh` exits before spawning `ovld` on every
+invocation. Treat the table below as the intended design this module retains, not live
+behavior — do not build against it assuming a channel exists at launch.
+
 The credential your adapter's CLI call presents is not something the adapter finds or derives:
 
 | Launch                 | Channel created at                       | Credential reaches the process via                  |
 | ---------------------- | ---------------------------------------- | --------------------------------------------------- |
-| Queued (runner)        | the `launching` transition               | owner-only local cache, keyed by channel id         |
-| Manual (`ovld launch`) | `POST /api/missions/:id/session-channel` | owner-only local cache, keyed by channel id         |
+| Queued (runner)        | the `launching` transition (not wired)   | owner-only local cache, keyed by channel id         |
+| Manual (`ovld launch`) | `POST /api/missions/:id/session-channel` (retired; returns 410) | owner-only local cache, keyed by channel id |
 | Virtual gateway        | the same preparation service             | `OVERLORD_SESSION_CHANNEL_TOKEN` in the environment |
 
 Only `OVERLORD_SESSION_CHANNEL_ID` is exported into the launch environment and the terminal

package/dist/connectors/README.md

@@ -86,10 +86,10 @@ contributor adding an adapter can tell signal from neglect. Source of truth for
 
 | Adapter       | Commands      | Hooks                                                 | Local MCP shim | MCP config        | Rules    | Native resume |
 | ------------- | ------------- | ----------------------------------------------------- | -------------- | ----------------- | -------- | ------------- |
-| `claude`      | `commands/`   | `hooks/hooks.json` (5 types)                          | —              | —                 | —        | —             |
+| `claude`      | `commands/`   | `hooks/hooks.json` (3 types)                          | —              | —                 | —        | —             |
 | `codex`       | —             | `.codex-plugin/hooks.json` + `scripts/*.sh` (2 types) | rendered       | `.mcp.json`       | —        | yes           |
 | `cursor`      | `commands/`   | `hooks/*.sh` (4 types)                                | rendered       | `mcp.json`        | `rules/` | —             |
-| `antigravity` | `skills/*.md` | `hooks.json` + `scripts/*.sh` (2 types)               | rendered       | `mcp_config.json` | —        | —             |
+| `antigravity` | `skills/*.md` | `hooks.json` (0 types — none shipped yet)             | rendered       | `mcp_config.json` | —        | —             |
 | `pi`          | —             | `extensions/overlord.ts` (input/tool/session events)  | —              | —                 | —        | yes           |
 | `opencode`    | —             | none — control plane, see below                       | —              | —                 | —        | —             |
 
@@ -103,6 +103,11 @@ Intentional omissions:
   yet start it automatically. A supervised execution target must supply that launch integration.
 - **`claude` ships no local MCP shim.** The Claude plugin reaches Overlord's hosted
   MCP server, so a local stdio bridge would duplicate it.
+- **`claude` ships a 4th hook script, `scripts/permission-hook.sh`, that is not registered in
+  `hooks/hooks.json`.** It is a declared hazard (`shipped-permission-hook-inert` in
+  `connectors/adapters/claude/CAPABILITIES.md`), not live behavior: it calls a
+  `ovld protocol permission-request` subcommand that does not exist and is structurally
+  incapable of returning a decision even once wired in.
 - **`pi` ships no shim, hooks directory, or commands.** PI integrates through one TypeScript
   extension. It has no native permission dialog; its optional `tool_call` decision interceptor
   is fail-open and requires explicit workspace and project opt-in.

package/dist/connectors/VERSION

@@ -1 +1 @@
-0.3.39
+0.3.41

package/dist/connectors/adapters/antigravity/conformance-manifest.yaml

@@ -1,4 +1,4 @@
-contractVersion: "129"
+contractVersion: "130"
 componentType: connector
 componentKey: antigravity
 label: "Antigravity Connector"

package/dist/connectors/adapters/antigravity/plugin.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://antigravity.google/schemas/v1/plugin.json",
   "name": "overlord",
-  "version": "0.3.39",
+  "version": "0.3.41",
   "description": "Chat-driven access to Overlord mission protocol workflows through skills and MCP.",
   "author": {
     "name": "Cooperativ",

package/dist/connectors/adapters/claude/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "overlord",
-  "version": "0.3.39",
+  "version": "0.3.41",
   "description": "Overlord mission protocol workflow for Claude Code — attach/update/ask/deliver, slash commands, and hook-driven activity capture.",
   "author": {
     "name": "Cooperativ",

package/dist/connectors/adapters/claude/conformance-manifest.yaml

@@ -1,4 +1,4 @@
-contractVersion: "129"
+contractVersion: "130"
 componentType: connector
 componentKey: claude
 label: "Claude Code Connector"

package/dist/connectors/adapters/codex/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "overlord",
-  "version": "0.3.39",
+  "version": "0.3.41",
   "description": "Chat-driven access to Overlord mission protocol workflows with hook-driven activity capture.",
   "author": {
     "name": "Cooperativ",

package/dist/connectors/adapters/codex/conformance-manifest.yaml

@@ -1,4 +1,4 @@
-contractVersion: "129"
+contractVersion: "130"
 componentType: connector
 componentKey: codex
 label: "Codex Connector"

package/dist/connectors/adapters/cursor/.cursor-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "overlord",
-  "version": "0.3.39",
+  "version": "0.3.41",
   "description": "Overlord mission protocol workflow for Cursor — attach/update/ask/deliver, slash commands, MCP bridge, and follow-up capture.",
   "author": {
     "name": "Cooperativ",

package/dist/connectors/adapters/cursor/conformance-manifest.yaml

@@ -1,4 +1,4 @@
-contractVersion: "129"
+contractVersion: "130"
 componentType: connector
 componentKey: cursor
 label: "Cursor Connector"

package/dist/connectors/adapters/opencode/conformance-manifest.yaml

@@ -1,4 +1,4 @@
-contractVersion: "129"
+contractVersion: "130"
 componentType: connector
 componentKey: opencode
 label: "OpenCode Connector"

package/dist/connectors/adapters/pi/conformance-manifest.yaml

@@ -1,4 +1,4 @@
-contractVersion: "129"
+contractVersion: "130"
 componentType: connector
 componentKey: pi
 label: "PI Connector"

package/dist/connectors/core/overlord-mission/SKILL.md

@@ -21,7 +21,7 @@ Use this mode when the prompt already contains a mission ID or explicitly says t
 2. The attach response prints JSON to stdout containing `session.sessionKey`. The CLI also persists this key automatically so subsequent `ovld protocol` commands in the same working directory resolve it without `--session-key`. If auto-resolution fails, pass `--session-key <sessionKey>` explicitly on every subsequent call.
 3. Treat the Overlord mission prompt as authoritative for the objective, constraints, and delivery target. Begin executing the current objective immediately after attach; do not wait for more instructions or ask for confirmation. This differs from `connect` or `load-context`, which only retrieve mission context and never imply the agent should act.
 4. Post updates while working: `ovld protocol update --session-key <sessionKey> --mission-id <mission_id> --summary "..." --phase execute`.
-   During long mechanical stretches with nothing meaningful to post, send `ovld protocol heartbeat --session-key <sessionKey> --mission-id <mission_id> [--phase execute] [--percent <0-100>] [--note "..."]` instead of an empty update.
+   During long mechanical stretches with nothing meaningful to post, send `ovld protocol heartbeat --session-key <sessionKey> --mission-id <mission_id> [--phase execute] [--note "..."]` instead of an empty update.
 5. Follow-up messages after the initial mission are captured automatically by the installed `UserPromptSubmit` hook and stay in discussion intent while the mission is in review. Do not post `user_follow_up` manually unless the hook is unavailable.
 6. If blocked, call `ovld protocol ask --session-key <sessionKey> --mission-id <mission_id> --question "..."` and stop.
 7. Deliver last with `ovld protocol deliver --session-key <sessionKey> --mission-id <mission_id> --summary "..."`. Normal delivery is summary-only; objective-ledger evidence syncs independently.
@@ -116,7 +116,7 @@ ovld protocol update --session-key <sessionKey> --mission-id $MISSION_ID --summa
 ### Heartbeat
 
 ```bash
-ovld protocol heartbeat --session-key <sessionKey> --mission-id $MISSION_ID --phase execute --percent 40 --note "Running the integration suite"
+ovld protocol heartbeat --session-key <sessionKey> --mission-id $MISSION_ID --phase execute --note "Running the integration suite"
 ```
 
 Use `heartbeat` for liveness pings and transient UI telemetry when you have no meaningful narrative summary to post. It updates the attached session without creating a mission event.
@@ -173,8 +173,9 @@ opened file, and never make a separate call just to declare paths.
 Artifacts, optional change rationales, and a delivery report remain current annotations when
 they add review value. Use `--payload-json` when the full delivery object fits comfortably inline
 (roughly under 8 KB). Larger inline `--*-json` values are rejected; stream them through the
-corresponding `--*-file -` flag. A real scratch path belongs under `.overlord/tmp` and should be
-removed after delivery.
+corresponding `--*-file -` flag. A real scratch path belongs under `$OVERLORD_TMPDIR` (this launch's private
+`.overlord/tmp/sessions/<id>/` directory); `deliver` removes that directory for you, and nothing
+else under `.overlord/tmp` is yours to delete.
 
 If `heartbeat` succeeds but `deliver` or `update` fails, the session is likely fine — retry with the corresponding `--*-file -` flag instead of inline `--*-json`.
 
@@ -292,7 +293,7 @@ ovld protocol create --agent <agent-identifier> --objectives-json '[{"objective"
 ```
 
 ```bash
-ovld protocol prompt --agent <agent-identifier> --objectives-json '[{"objective":"Implement feature X"}]' --priority medium
+ovld protocol prompt --agent <agent-identifier> --objectives-json '[{"objective":"Implement feature X"}]'
 ```
 
 ```bash
@@ -303,17 +304,6 @@ ovld protocol add-objectives --mission-id 1:899 --objectives-json '[{"objective"
 ovld protocol update-objective --objective-id <objective-uuid> --auto-advance
 ```
 
-#### Local Durability For New Missions
-
-`create`, `prompt`, `add-objectives`, and `record-work` save the objective/mission text to a local draft (`~/.overlord/pending-missions/`) **before** sending it, and delete that draft only once the server confirms the write. If the network drops mid-call, your text is never lost — the failure message points at the saved file. Manage outstanding drafts with `pending-missions`:
-
-```bash
-ovld protocol pending-missions               # list drafts the server never confirmed
-ovld protocol pending-missions --retry <id>  # re-send a saved draft; clears it on success
-ovld protocol pending-missions --clear <id>  # delete one draft after confirming it landed
-ovld protocol pending-missions --clear-all   # delete every draft
-```
-
 To inspect project resolution explicitly:
 
 ```bash
@@ -342,11 +332,17 @@ ovld protocol create-project --name "Acme Web" --directory /path/to/repo
 ovld protocol create-project --name "Acme Web" --no-directory
 ```
 
-`ovld create-project` is a friendly top-level alias for `ovld protocol create-project`.
-When a directory is registered the command also writes `.overlord/project.json` so
-future cwd-based resolution finds the project. Pass `--organization-id <id>` to create
-in a specific organization (defaults to your membership); `--color <#rrggbb>` sets the
-project color.
+`ovld create-project --name "<name>"` is the top-level convenience command; it accepts
+only `--name`, `--directory`/`--no-directory`, and `--json`, and always creates the
+project in your resolved workspace. When a directory is registered it also writes
+`.overlord/project.json` so future cwd-based resolution finds the project.
+
+`ovld protocol create-project --name "<name>"` is the parentless Protocol surface;
+unlike the top-level command it does not link a local directory. Pass
+`--workspace-id <id|slug|name>` to choose a workspace when you belong to more than one
+(omitting it returns `workspace_selection_required` with your choices); optional
+`--description <text>` and `--slug <text>` are also accepted. There is no `--color`
+flag on either surface.
 
 #### Resolving the project ID when you don't have one
 

package/dist/connectors/core/overlord-mission/reference/cli.md

@@ -41,7 +41,7 @@ ovld protocol update --session-key <sessionKey> --mission-id $MISSION_ID --summa
 ## Heartbeat
 
 ```bash
-ovld protocol heartbeat --session-key <sessionKey> --mission-id $MISSION_ID --phase execute --percent 40 --note "Running the integration suite"
+ovld protocol heartbeat --session-key <sessionKey> --mission-id $MISSION_ID --phase execute --note "Running the integration suite"
 ```
 
 Use `heartbeat` for liveness pings and transient UI telemetry when you have no meaningful narrative summary to post. It updates the attached session without creating a mission event.
@@ -189,7 +189,7 @@ ovld protocol create --agent <agent-identifier> --objectives-json '[{"objective"
 ```
 
 ```bash
-ovld protocol prompt --agent <agent-identifier> --objectives-json '[{"objective":"Implement feature X"}]' --priority medium
+ovld protocol prompt --agent <agent-identifier> --objectives-json '[{"objective":"Implement feature X"}]'
 ```
 
 ```bash
@@ -223,17 +223,6 @@ EOF
 The full submission format, including the `record-work`-only `changedFiles` field and MCP
 equivalent, is in [record-work.md](record-work.md).
 
-### Local Durability For New Missions
-
-`create`, `prompt`, `add-objectives`, and `record-work` save the objective/mission text to a local draft (`~/.overlord/pending-missions/`) **before** sending it, and delete that draft only once the server confirms the write. If the network drops mid-call, your text is never lost — the failure message points at the saved file. Manage outstanding drafts with `pending-missions`:
-
-```bash
-ovld protocol pending-missions               # list drafts the server never confirmed
-ovld protocol pending-missions --retry <id>  # re-send a saved draft; clears it on success
-ovld protocol pending-missions --clear <id>  # delete one draft after confirming it landed
-ovld protocol pending-missions --clear-all   # delete every draft
-```
-
 To inspect project resolution explicitly:
 
 ```bash
@@ -262,11 +251,17 @@ ovld protocol create-project --name "Acme Web" --directory /path/to/repo
 ovld protocol create-project --name "Acme Web" --no-directory
 ```
 
-`ovld create-project` is a friendly top-level alias for `ovld protocol create-project`.
-When a directory is registered the command also writes `.overlord/project.json` so
-future cwd-based resolution finds the project. Pass `--organization-id <id>` to create
-in a specific organization (defaults to your membership); `--color <#rrggbb>` sets the
-project color.
+`ovld create-project --name "<name>"` is the top-level convenience command; it accepts
+only `--name`, `--directory`/`--no-directory`, and `--json`, and always creates the
+project in your resolved workspace. When a directory is registered it also writes
+`.overlord/project.json` so future cwd-based resolution finds the project.
+
+`ovld protocol create-project --name "<name>"` is the parentless Protocol surface;
+unlike the top-level command it does not link a local directory. Pass
+`--workspace-id <id|slug|name>` to choose a workspace when you belong to more than one
+(omitting it returns `workspace_selection_required` with your choices); optional
+`--description <text>` and `--slug <text>` are also accepted. There is no `--color`
+flag on either surface.
 
 ### Resolving the project ID when you don't have one
 

package/dist/connectors/core/overlord-mission/reference/mcp.md

@@ -10,6 +10,7 @@
 - `overlord_record_work` (hosted MCP) and `ovld protocol record-work` (CLI) record already-completed chat work as a single review-column mission with a completed objective, explicitly supplied changed files, optional rationales, and the standard Gemini delivery summary — no `attach`/`deliver`. Hosted MCP requires `projectId` (it never chooses a project implicitly); the CLI resolves the project from the working directory or `--project-id`. The submission format is in [record-work.md](record-work.md). The route is `POST /api/protocol/record-work`.
 - `overlord_add_artifact` (hosted MCP) and `ovld protocol add-artifact` (CLI) create a mission artifact during a turn without delivering. Required: `type` / `--type`, `label` / `--label`, mission scope (`missionId` / `--mission-id`, or an objective display id as `objectiveId` / `--objective-id`), plus at least one of `contentText` / `--content-text`/`--content-text-file` or `externalUrl` / `--external-url`. Optional `sessionKey` / `--session-key` stamps session/objective provenance; `objectiveId` / `--objective-id` does the same when there is no live session, and the session key wins if both are given. Same service as REST `POST /api/missions/:id/artifacts`. The route is `POST /api/protocol/add-artifact`.
 - `overlord_update_artifact` (hosted MCP) and `ovld protocol update-artifact` (CLI) revise an existing mission artifact's label, Markdown content, and/or HTTP(S) URL in place. Required: `artifactId` / `--artifact-id`, `expectedRevision` / `--expected-revision`, and mission scope (`missionId` / `--mission-id`, or an objective display id as `objectiveId` / `--objective-id`). Provide at least one editable field. No session key is required. Same service as REST `PATCH /api/missions/:id/artifacts/:artifactId`. The route is `POST /api/protocol/update-artifact`.
+- `overlord_delete_missions` / `overlord_delete_objectives` (hosted MCP) and `ovld protocol delete-missions` / `delete-objectives` (CLI) bulk soft-delete one to 100 mission or objective UUID/display-id references atomically. They require `confirm: true` / `--confirm`; before setting it, agents must show the resolved list to the user and receive affirmative approval. A missing, duplicate, deleted, or unauthorized target rejects the entire request. Mission deletion includes live child objectives; objective deletion removes live Run Queue membership.
 - Run Queue MCP stays a thin Protocol surface: `overlord_list_run_queues` reads compact/full state; `overlord_reorder_run_queue` atomically replaces one queue's complete entry order; and `overlord_queue_objective` accepts `objectiveId`, optional `queue` / `after` / `front` / `position`, and `remove`. `overlord_manage_run_queue` creates, updates, deletes, and reorders queue definitions. Entry operations use `execution_request:create`; definition operations use `project:update`, so they require a full-scope token. Instruction text remains editable only in `draft` or `future` state. The routes are `POST /api/protocol/run-queue`, `reorder-run-queue`, `queue-objective`, `dequeue-objective`, `create-run-queue`, `update-run-queue`, `delete-run-queue`, `reorder-project-run-queues`, and `update-objective`.
 - Objective attachment tools follow the `<verb>_<noun>` MCP naming: `list_attachments`, `prepare_attachment_upload`, `finalize_attachment_upload`, `get_attachment_download_url`, `upload_attachment_file`. CLI commands use `attachment-*`. `mission_id` / `missionId` is optional for attachment calls when the objective or attachment id lets the server derive mission scope.
 - "Artifacts" may be published mid-turn with `add-artifact` / `overlord_add_artifact`, revised with `update-artifact` / `overlord_update_artifact`, or attached at `deliver` time (next_steps, test_results, migration, decision, note, url) — not user-uploaded files. Files attached by users live on objectives via the attachment tools above.

package/dist/connectors/core/scripts/overlord-mcp.mjs

@@ -110,16 +110,14 @@ function compactSearchResponse(response) {
     results: value.results.map(result => ({
       ...result,
       matches: Array.isArray(result.matches)
-        ? result.matches
-            .slice(0, 2)
-            .map(match => {
-              const compactMatch = { ...match };
-              delete compactMatch.snippet;
-              delete compactMatch.matchedTerms;
-              delete compactMatch.createdAt;
-              delete compactMatch.updatedAt;
-              return compactMatch;
-            })
+        ? result.matches.slice(0, 2).map(match => {
+            const compactMatch = { ...match };
+            delete compactMatch.snippet;
+            delete compactMatch.matchedTerms;
+            delete compactMatch.createdAt;
+            delete compactMatch.updatedAt;
+            return compactMatch;
+          })
         : result.matches
     }))
   };
@@ -312,7 +310,7 @@ const tools = [
     name: 'overlord_list_deliveries',
     title: 'List mission deliveries',
     description:
-      'Use this to read a mission\'s delivered work after locating it. Returns newest-first normalized delivery records including summary, verification, follow-up notes, and authoritative delivery evidence; it never exposes raw delivery payloads.',
+      "Use this to read a mission's delivered work after locating it. Returns newest-first normalized delivery records including summary, verification, follow-up notes, and authoritative delivery evidence; it never exposes raw delivery payloads.",
     inputSchema: objectSchema({
       missionId: stringProperty(
         'Mission UUID or workspace display id. Optional when objectiveId is a display id such as coo:756.k7xm, which already names its mission.'
@@ -336,7 +334,7 @@ const tools = [
           'Optional reasoning-effort setting to snapshot onto the execution request.'
         ),
         executionTargetId: stringProperty(
-          'Optional eligible execution target id. Omit to use the project\'s normal target selection.'
+          "Optional eligible execution target id. Omit to use the project's normal target selection."
         )
       },
       ['objectiveId', 'agent']
@@ -407,6 +405,44 @@ const tools = [
       ['objectiveId']
     )
   },
+  {
+    name: 'overlord_delete_missions',
+    title: 'Delete missions',
+    description:
+      'Use this only after showing the resolved mission list to the user and receiving an affirmative response. Soft-deletes one to 100 missions atomically; any missing, duplicate, deleted, or unauthorized target rejects the entire request. Deleting a mission also soft-deletes its live objectives.',
+    inputSchema: objectSchema(
+      {
+        missionIds: {
+          type: 'array',
+          description: 'One to 100 unique mission UUIDs or display IDs to delete.',
+          items: stringProperty('Mission UUID or workspace display ID.')
+        },
+        confirm: booleanProperty(
+          'Must be true only after the user has reviewed the resolved mission list and affirmatively approved deletion.'
+        )
+      },
+      ['missionIds', 'confirm']
+    )
+  },
+  {
+    name: 'overlord_delete_objectives',
+    title: 'Delete objectives',
+    description:
+      'Use this only after showing the resolved objective list to the user and receiving an affirmative response. Soft-deletes one to 100 objectives atomically; any missing, duplicate, deleted, or unauthorized target rejects the entire request. Deleting an objective also removes its live Run Queue membership.',
+    inputSchema: objectSchema(
+      {
+        objectiveIds: {
+          type: 'array',
+          description: 'One to 100 unique objective UUIDs or display IDs to delete.',
+          items: stringProperty('Objective UUID or display ID.')
+        },
+        confirm: booleanProperty(
+          'Must be true only after the user has reviewed the resolved objective list and affirmatively approved deletion.'
+        )
+      },
+      ['objectiveIds', 'confirm']
+    )
+  },
   {
     name: 'overlord_list_run_queues',
     title: 'List Run Queues',
@@ -485,12 +521,14 @@ const tools = [
           type: 'string',
           enum: ['create', 'update', 'delete', 'reorder_queues', 'retry_entry'],
           description:
-            'create a queue, update (rename/pause/resume) one, delete one, reorder the queue definitions, or retry_entry to clear one held entry\'s hold and attempt budget so the dispatcher tries it again.'
+            "create a queue, update (rename/pause/resume) one, delete one, reorder the queue definitions, or retry_entry to clear one held entry's hold and attempt budget so the dispatcher tries it again."
         },
         projectId: stringProperty(
           'Project UUID, slug, or name. Required for create and reorder_queues.'
         ),
-        queue: stringProperty('Run Queue UUID or unambiguous name. Required for update and delete.'),
+        queue: stringProperty(
+          'Run Queue UUID or unambiguous name. Required for update and delete.'
+        ),
         name: stringProperty('New queue name for create, or the replacement name for update.'),
         paused: booleanProperty(
           'update only: true pauses dispatching from the queue, false resumes it.'
@@ -512,7 +550,7 @@ const tools = [
         }
       },
       ['action']
-    ),
+    )
   },
   {
     name: 'overlord_attach_session',
@@ -862,6 +900,33 @@ async function callOverlordTool(name, args) {
       ...(hasInstructionText ? { 'instruction-text': requiredString(args, 'instructionText') } : {})
     });
   }
+  if (name === 'overlord_delete_missions') {
+    if (!Array.isArray(args.missionIds) || !args.missionIds.every(id => typeof id === 'string')) {
+      throw new Error('missionIds must be an array of strings');
+    }
+    if (args.confirm !== true) {
+      throw new Error('confirm must be true after the user has affirmatively approved deletion');
+    }
+    return runProtocol('delete-missions', {
+      'mission-ids-json': args.missionIds,
+      confirm: true
+    });
+  }
+  if (name === 'overlord_delete_objectives') {
+    if (
+      !Array.isArray(args.objectiveIds) ||
+      !args.objectiveIds.every(id => typeof id === 'string')
+    ) {
+      throw new Error('objectiveIds must be an array of strings');
+    }
+    if (args.confirm !== true) {
+      throw new Error('confirm must be true after the user has affirmatively approved deletion');
+    }
+    return runProtocol('delete-objectives', {
+      'objective-ids-json': args.objectiveIds,
+      confirm: true
+    });
+  }
   if (name === 'overlord_list_run_queues') {
     const detail = optionalString(args, 'detail') || 'compact';
     if (detail !== 'compact' && detail !== 'full') {
@@ -1110,7 +1175,7 @@ process.stdin.on('data', async chunk => {
         result: {
           protocolVersion: PROTOCOL_VERSION,
           capabilities: { tools: { listChanged: false } },
-          serverInfo: { name: 'overlord-__OVERLORD_ADAPTER_KEY__', version: '0.3.39' }
+          serverInfo: { name: 'overlord-__OVERLORD_ADAPTER_KEY__', version: '0.3.41' }
         }
       });
       continue;

package/dist/connectors/docs/05-connectors-and-agent-plugins.md

@@ -236,6 +236,14 @@ Requirements:
 - Degrades gracefully without a web app open: the harness's own fallback
   decision (allow/deny/ask) applies when no one answers in time.
 
+As of the Latch v2 cutover, no shipped launch path currently supplies the session-channel
+credential this mechanism depends on (`createSessionChannel` has no production callers), so
+"blocks the harness's tool call until answered" does not happen end-to-end for any existing
+adapter today — do not assume a new connector can exercise this path against a real Overlord
+deployment without first restoring that wiring. The one decision kind that is still remotely
+answerable — a blocking **question** — goes through the mission protocol's `ask`/resolve flow
+and a live Latch terminal session, not through this session-channel mechanism.
+
 ### Stop Hook
 
 Future requirement: