@bridge_gpt/mcp-server 0.2.34 → 0.2.36

package/docs/install/mcp-tool-integrations.md

@@ -48,26 +48,31 @@ marked **— none**.
 
 ## Profiles (a form of BLOCK)
 
-The server registers **90 tools** total, gated by `BRIDGE_MCP_PROFILE`
-(comma-separated group list, resolved once at startup by
-`mcp_server/src/mcp-profile.ts`). A tool outside
-the active profile is **not registered at all** — an absolute block until the group is
+Tool registration is gated by `BRIDGE_MCP_PROFILE` (comma-separated group list,
+resolved once at startup by `mcp_server/src/mcp-profile.ts`). A tool outside the
+active profile is **not registered at all** — an absolute block until the group is
 activated. `core` is always implicitly included.
 
-| Profile | Tools | Registered when |
-|---|---|---|
-| `core` | **57** | always |
-| `pipeline-authoring` | **+5** | `BRIDGE_MCP_PROFILE` includes `pipeline-authoring` or `full` |
-| `conductor` | **+8** | includes `conductor` or `full` |
-| `sfcc` | **+20** (2 diagnostics ship in `core`; 18 gated) | includes `sfcc` or `full` |
+| Profile | Registered when |
+|---|---|
+| `core` | always |
+| `pipeline-authoring` | `BRIDGE_MCP_PROFILE` includes `pipeline-authoring` or `full` |
+| `conductor` | includes `conductor` or `full` |
+| `sfcc` | includes `sfcc` or `full` (2 diagnostics ship in `core`; the rest are gated) |
+
+The `pipeline-authoring` tools live in `index.ts` alongside core, gated by
+`ACTIVE_GROUPS.has("pipeline-authoring")`.
 
-(57 core + 5 + 8 + 20 = 90. The 5 pipeline-authoring tools live in `index.ts` alongside
-core, gated by `ACTIVE_GROUPS.has("pipeline-authoring")`, so `index.ts` holds 62
-`registerTool` calls.)
+> **This document deliberately states no tool counts.** Counts drift on every
+> ticket that adds or moves a tool, and nothing here is generated, so a number
+> written down is a number that goes stale. For the authoritative surface, read
+> `api/library/config/mcp_tool_catalog.json` (the committed catalog, kept in sync
+> by `scripts/sync_mcp_tool_catalog.py --check`) or call `tools/list` against a
+> running server. The per-tool rows below are what this document is for.
 
 ---
 
-## Core profile (57)
+## Core profile
 
 ### Connectivity, config & setup — no integration deps
 
@@ -78,6 +83,8 @@ These are never gated (`NEVER_GATED_ROUTE_KEYS`) or run entirely locally.
 | `ping` | — none |
 | `get_project_standards` | — none |
 | `get_my_role` | — none |
+| `invite_member` | — none (mints a scoped key in the Bridge DB; admin-only) |
+| `resolve_target_status` | — none (status resolution over the repo's configured mapping) |
 | `config_field` (get/update/list) | — none |
 | `get_install_manifest` | — none |
 | `apply_install_manifest` | — none |
@@ -87,6 +94,7 @@ These are never gated (`NEVER_GATED_ROUTE_KEYS`) or run entirely locally.
 | `track_ticket` | — none (Bridge DB tracking record) |
 | `update_ticket_state` | — none (Bridge DB) |
 | `get_ticket_state` | — none (Bridge DB) |
+| `get_ticket_state_tree` | — none (Bridge DB; live projection over existing rows) |
 | `get_ticket_model_tier` | — none (Bridge DB difficulty) |
 | `get_pipeline_recipe` | — none (`LOCAL`, bundled recipes) |
 | `generate_decision_page` | — none (`LOCAL` HTML) |
@@ -186,7 +194,7 @@ These orchestrate an idea → ticket → review → start chain, so they inherit
 
 ---
 
-## `pipeline-authoring` profile (5)
+## `pipeline-authoring` profile
 
 Profile-gated **[BLOCK]** to register at all. `run_pipeline` / `resume_pipeline`
 additionally inherit the dependencies of whatever steps the invoked pipeline runs
@@ -203,7 +211,7 @@ Run state is persisted via the Bridge API.
 
 ---
 
-## `conductor` profile (8)
+## `conductor` profile
 
 Profile-gated **[BLOCK]** to register. Most operate on the **local** conductor ledger
 (`~/.config/bridge/events.db`), so beyond the profile gate they have no remote
@@ -222,7 +230,7 @@ integration dependency.
 
 ---
 
-## `sfcc` group (20)
+## `sfcc` group
 
 ### Always-on diagnostics (ship in `core`)
 
@@ -231,13 +239,13 @@ integration dependency.
 | `sfcc_setup_status` | — none (its purpose is to report which SFCC prerequisites are missing) |
 | `check_permissions` | SFCC OCAPI **[BLOCK]** (probes OCAPI; reports not-OK without access / version config) |
 
-### Gated behind the `sfcc` group (18)
+### Gated behind the `sfcc` group
 
 All are profile-gated **[BLOCK]** to register, and all require SFCC OCAPI creds +
 `version` config + a reachable sandbox → SFCC OCAPI **[BLOCK]**. Writes are
 sandbox-only and destructive.
 
-**Reads (8):**
+**Reads:**
 
 | Tool | Dependencies (class) |
 |---|---|
@@ -264,7 +272,7 @@ sandbox-only and destructive.
 | `custom_object_definition_attribute_update` | Profile-gated **[BLOCK]** · SFCC OCAPI write **[BLOCK]** |
 | `site_preference_values_set` | Profile-gated **[BLOCK]** · SFCC OCAPI write **[BLOCK]** |
 
-**Logs (1):**
+**Logs:**
 
 | Tool | Dependencies (class) |
 |---|---|
@@ -283,7 +291,7 @@ sandbox-only and destructive.
 | **SFCC OCAPI** | `check_permissions` + all 16 SFCC read/write tools | — |
 | **SFCC WebDAV logs** | `sfcc_log_query` | — |
 | **Deep-research flag** | `request_deep_research`, `get_deep_research` | — |
-| **Profile gating** | all `pipeline-authoring` (5), `conductor` (8), gated `sfcc` (18) tools | — |
+| **Profile gating** | all `pipeline-authoring`, `conductor`, and gated `sfcc` tools | — |
 
 ## Notes & caveats
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@bridge_gpt/mcp-server",
-  "version": "0.2.34",
+  "version": "0.2.36",
   "description": "Bridge API MCP server — exposes Jira endpoints as MCP tools for Claude Code agents",
   "license": "MIT",
   "type": "module",
@@ -20,13 +20,16 @@
     "docs/",
     "LICENSE"
   ],
+  "//test": "The `test` script delegates to the bounded launcher (scripts/run-unit-tests.js → src/run-unit-tests-launcher.ts, BAPI-683), which discovers build/*.test.js at runtime and spawns them in size-bounded batches so the argv never exceeds Windows' ~8191-char command-line limit. Two determinism rules from the pre-launcher script are preserved INSIDE the launcher: (1) multi-file batches run WITHOUT --test-force-exit, because that flag makes node:test exit before aggregating per-subprocess summaries and the tally becomes nondeterministic (failures still surface and still set exit 1, but the count cannot be trusted as a completeness signal); (2) build/secret-safety.test.js leaks a handle that holds the event loop ~60s after its tests finish (they run in ~147ms), so the launcher quarantines it into its own single-file --test-force-exit invocation (FORCE_EXIT_QUARANTINE in src/run-unit-tests-launcher.ts) where the tally stays exact. Do NOT add --test-force-exit to batched invocations, and do NOT remove the quarantine, until the underlying handle leak is fixed.",
   "scripts": {
-    "build": "node scripts/bundle-version.js && node scripts/bundle-readme.js && node scripts/bundle-pipelines.js && node scripts/bundle-commands.js && node scripts/bundle-agents.js && node scripts/bundle-docs.js && tsc && node scripts/bundle-esbuild.js",
+    "build": "node scripts/bundle-version.js && node scripts/bundle-readme.js && node scripts/bundle-pipelines.js && node scripts/bundle-commands.js && node scripts/bundle-agents.js && node scripts/bundle-docs.js && tsc && node scripts/sync-agent-mirrors.js && node scripts/bundle-esbuild.js",
     "check:version-generated": "node scripts/bundle-version.js && node scripts/check-version-generated.js",
     "postbuild": "node scripts/prepend-shebang.cjs",
     "start": "node build/index.js",
-    "test": "node --test --test-force-exit build/pipeline-utils.test.js build/backend-warnings.test.js build/update-check.test.js build/cli-upgrade.test.js build/decision-page-schema.test.js build/decision-page-template.test.js build/bundle-pipelines.test.js build/instructions-contract.test.js build/pipeline-orchestrator-persistence.test.js build/pipeline-orchestrator-execution.test.js build/pipeline-orchestrator-integration.test.js build/index-static.test.js build/config-fields.static.test.js build/request-brainstorm.static.test.js build/execute-plan-instructions.static.test.js build/index-resolvers.test.js build/index-project-root.test.js build/index-pipelines.test.js build/index.test.js build/bridge-config.test.js build/credential-store.test.js build/agent-config-credential-migration.test.js build/mcp-invoke.test.js build/mcp-provisioning.test.js build/mcp-server-invocation.test.js build/third-party-mcp-targets.test.js build/git-ignore-utils.test.js build/command-provisioning.test.js build/command-assets-doctor.test.js build/credential-materialization.test.js build/mcp-registration-doctor.test.js build/mcp-host-targets.test.js build/mcp-install-state.test.js build/mcp-host-config.test.js build/secret-safety.test.js build/base-ref.test.js build/pr-base-contract.test.js build/start-tickets.test.js build/review-tickets.test.js build/start-tickets-base-branch.test.js build/agent-registry.test.js build/agent-registry.model-routing.test.js build/start-tickets.shell-model-routing.test.js build/start-tickets.bridge-api-model-routing.test.js build/start-tickets.tier-fetch-model-routing.test.js build/start-tickets.resolve-model-routing.test.js build/start-tickets.orchestrate-model-routing.test.js build/start-tickets.routing-diagnostics.test.js build/start-tickets-repo.test.js build/start-tickets-credential-invariants.static.test.js build/credentials-cli.test.js build/start-tickets-prereqs.test.js build/regression-check.test.js build/doctor.test.js build/install-doctor.test.js build/install-bridge.test.js build/install-bridge-invite.test.js build/install-bridge-prompt.test.js build/install-bridge-tools.test.js build/install-bridge-join-static.test.js build/init.test.js build/init-docs.test.js build/resolveUploadAttachment.test.js build/package-static.test.js build/chain-utils.test.js build/chain-orchestrator.test.js build/scheduler-backends/types.test.js build/scheduler-backends/escaping.test.js build/scheduler-backends/launchd.test.js build/scheduler-backends/task-scheduler.test.js build/scheduler-backends/systemd-user.test.js build/scheduler-backends/at-fallback.test.js build/scheduler-backends/index.test.js build/command-catalog.test.js build/scheduled-prompt.test.js build/agent-launchers/claude.test.js build/agent-launchers/cursor.test.js build/agent-launchers/index.test.js build/schedule-store.test.js build/schedule-run.test.js build/setup-epic.test.js build/conductor-bundle-artifacts.test.js build/conductor-bundle-cli.test.js build/agent-capabilities/cli.test.js build/agent-capabilities/runner.test.js build/agent-capabilities/probes.test.js build/agent-capabilities/probe-context.test.js build/agent-capabilities/reporter.test.js build/conductor/taxonomy-and-errors.test.js build/conductor/github-mergeability.test.js build/conductor/merge-conflict-routing.test.js build/conductor/redaction-normalization.test.js build/conductor/claude-hook.test.js build/conductor/git-ci-types.test.js build/conductor/done-gate.test.js build/conductor/git-ci-taxonomy-payload.test.js build/conductor/bridge-api-client.test.js build/conductor/plan.test.js build/conductor/producer-ledger.test.js build/conductor/spec-review-producer.test.js build/conductor/git-producer.test.js build/conductor/git-hooks.test.js build/conductor/store-migration.test.js build/conductor/pr-discovery.test.js build/conductor/pr-ci-producer.test.js build/conductor/pr-review-producer.test.js build/conductor/doctor.test.js build/conductor/deny-enforcement-preflight.test.js build/conductor/errors.test.js build/conductor/store.test.js build/conductor/index-poll-ci-producer.test.js build/start-tickets-conductor.test.js build/start-tickets-conductor.spawn.test.js build/start-tickets-conductor-node-propagation.test.js build/start-tickets.non-mutating-base.test.js build/start-tickets-live-source-guard.test.js build/conductor/worker-ledger-cli.test.js build/conductor/supervisor-config.test.js build/conductor/supervisor-ledger.test.js build/conductor/supervisor-state-reducer.test.js build/conductor/supervisor-housekeeping-projection.test.js build/conductor/supervisor-escalation.test.js build/conductor/supervisor-retired-judgment.static.test.js build/conductor/supervisor-runtime.test.js build/conductor/supervisor-store-projection.test.js build/conductor/supervisor-cli.test.js build/conductor/supervisor-start-tickets.test.js build/conductor/supervisor-message-relay.test.js build/conductor/supervisor-state-message-events.test.js build/conductor/store-message-relay.test.js build/start-tickets-message-relay.test.js build/conductor/event-accessors.test.js build/conductor/merge-ledger.test.js build/conductor/deterministic-completion.static.test.js build/conductor/deterministic-completion.integration.test.js build/conductor/local-merge.test.js build/conductor/local-merge.static.test.js build/conductor/supervisor-merge.test.js build/conductor/bridge-api-merge-client.test.js build/conductor/bridge-api-epic-client.test.js build/conductor/supervisor-merge-runtime-state.test.js build/conductor/epic-state.test.js build/executor/claude-settings.test.js build/executor/cli.test.js build/executor/credentials.test.js build/executor/echo-acceptance.test.js build/executor/env.test.js build/executor/base-branch.test.js build/executor/heartbeat.test.js build/executor/http-client.test.js build/executor/job-runner.test.js build/executor/job-runner.payload-timeout.test.js build/executor/worker-finalization.test.js build/executor/job-runner.static.test.js build/executor/prompt-spec.test.js build/executor/job-types.test.js build/executor/recovery-job.static.test.js build/executor/observation.test.js build/executor/permissions.test.js build/executor/preflight.test.js build/executor/process.test.js build/executor/results.test.js build/executor/runner.test.js build/executor/terminal-mutation.test.js build/executor/worker-command.test.js build/executor/worktree-core.test.js build/executor/worktree.test.js build/executor/job-errors.test.js build/executor/worktree-inspection.test.js build/executor/resume-pre-spawn.test.js build/executor/verdict-artifact.test.js build/executor/worker-log.test.js build/executor/job-log-registry.test.js build/executor/viewer-tabs.test.js build/executor/watch-cli.test.js build/executor/merge-job.test.js build/executor/job-runner.merge.test.js build/executor/worktree-gc.test.js build/executor/runner.job-behaviors.test.js build/conductor/cli-freeze.test.js build/conductor/epic-reconcile.test.js build/conductor/epic-runtime.test.js build/conductor/epic-tick-sequence.test.js build/conductor/epic-runtime-post-action.test.js build/conductor/file-scope-guard.test.js build/conductor/file-scope-guard.integration.test.js build/mcp-profile.test.js build/mcp-profile-registration.test.js build/env-flags.test.js build/bridge-api-urls.test.js build/tool-surface-gating.test.js build/tools-budget.test.js build/visual-diff-worker.test.js build/visual-diff.test.js build/estimate-epic.test.js build/integration/measure-tools.test.js build/sfcc/config.test.js build/sfcc/ocapi-shape.test.js build/sfcc/output.test.js build/sfcc/credentials.test.js build/sfcc/ocapi-write-faults.test.js build/sfcc/write-guard.test.js build/sfcc/write-grants.test.js build/sfcc/write-result.test.js build/sfcc/writes.test.js build/sfcc/writes-system-object-payloads.test.js build/sfcc/writes-payloads.test.js build/sfcc/log-gate.test.js build/sfcc/log-query.test.js && node --experimental-test-module-mocks --test --test-force-exit build/index-heavy-read-truncation.test.js build/index-artifacts.test.js build/automation-progress.test.js build/recovery-formatting.test.js build/wait-for-result.test.js build/ticket-wait-recovery.test.js build/council-wait-recovery.test.js build/index.review-rounds.test.js build/index-brainstorm-filenames.test.js build/index-output-path.test.js build/index-generate-decision-page.test.js build/index-generate-decision-page.integration.test.js build/visual-diff.registration.test.js build/visual-diff.attachment-adapter.test.js build/attachment-download.test.js build/attachment-upload.test.js build/conductor/paths.test.js build/conductor/store-lifecycle.test.js build/conductor/store-queries.test.js build/conductor/tools.test.js build/conductor/cli.test.js build/conductor/security-regressions.test.js build/conductor/git-inspection.test.js build/conductor/tools-done-gate.test.js build/conductor/pr-ci-producer-emit-seam.test.js build/conductor/cli-git-hooks.test.js build/sfcc/client.test.js build/sfcc/tool-wrapper.test.js build/sfcc/setup-status.test.js build/sfcc/permissions.test.js build/sfcc/register.test.js build/sfcc/reads-system-object.test.js build/sfcc/reads-custom-object-def.test.js build/sfcc/reads-site-preference.test.js build/sfcc/writes-system-object.test.js build/sfcc/writes-custom-object-def.test.js build/sfcc/writes-site-preference.test.js build/connect-github.test.js build/connect-github-api.test.js build/connect-github-handoff.test.js build/connect-github-dispatch.static.test.js",
-    "test:integration": "node --test build/integration/refresh-main.integration.test.js build/integration/command-provisioning.integration.test.js build/integration/start-tickets.integration.test.js build/integration/start-tickets-tier-handoff.integration.test.js build/integration/doctor.integration.test.js build/integration/agent-capabilities.integration.test.js build/integration/conductor-producer.integration.test.js build/integration/conductor-message-relay.integration.test.js build/integration/executor-http-runner.integration.test.js build/integration/executor-job-behaviors.integration.test.js build/integration/executor-recovery-jobs.integration.test.js build/integration/executor-spec-review-prompt.integration.test.js build/integration/resume-pre-spawn.git.integration.test.js build/integration/worker-finalization-origin.integration.test.js build/integration/post-remediation-merge-ci-wait.integration.test.js build/integration/executor-merge-supervision.integration.test.js build/integration/attachment-binary-roundtrip.integration.test.js build/integration/dependent-ticket-fresh-base.integration.test.js build/integration/execute-plan-instructions.integration.test.js build/integration/conductor-bundle-artifacts.integration.test.js build/integration/install-bridge-repo-resolution.integration.test.js build/integration/capability-report-contract.integration.test.js build/integration/request-brainstorm-general.integration.test.js build/integration/request-council-trigger-drop.integration.test.js build/integration/install-bridge-onboarding-launch.integration.test.js build/integration/install-bridge-failure-guards.integration.test.js build/integration/learn-repository-pipeline.integration.test.js build/integration/visual-diff-mcp.integration.test.js",
+    "test": "npm run test:normal && npm run test:module-mocks",
+    "test:normal": "node scripts/run-unit-tests.js normal",
+    "test:module-mocks": "node scripts/run-unit-tests.js module-mocks",
+    "test:integration": "node --test build/integration/refresh-main.integration.test.js build/integration/command-provisioning.integration.test.js build/integration/start-tickets.integration.test.js build/integration/start-tickets-tier-handoff.integration.test.js build/integration/doctor.integration.test.js build/integration/agent-capabilities.integration.test.js build/integration/conductor-producer.integration.test.js build/integration/conductor-message-relay.integration.test.js build/integration/executor-http-runner.integration.test.js build/integration/executor-job-behaviors.integration.test.js build/integration/executor-recovery-jobs.integration.test.js build/integration/executor-spec-review-prompt.integration.test.js build/integration/resume-pre-spawn.git.integration.test.js build/integration/worker-finalization-origin.integration.test.js build/integration/post-remediation-merge-ci-wait.integration.test.js build/integration/executor-merge-supervision.integration.test.js build/integration/attachment-binary-roundtrip.integration.test.js build/integration/dependent-ticket-fresh-base.integration.test.js build/integration/execute-plan-instructions.integration.test.js build/integration/implement-ticket-finalization-order.integration.test.js build/integration/conductor-bundle-artifacts.integration.test.js build/integration/install-bridge-repo-resolution.integration.test.js build/integration/capability-report-contract.integration.test.js build/integration/request-brainstorm-general.integration.test.js build/integration/request-council-trigger-drop.integration.test.js build/integration/install-bridge-onboarding-launch.integration.test.js build/integration/install-bridge-conductor.integration.test.js build/integration/install-bridge-failure-guards.integration.test.js build/integration/learn-repository-pipeline.integration.test.js build/integration/visual-diff-mcp.integration.test.js build/integration/executor-mcp-provisioning.integration.test.js build/integration/serve-stdio.integration.test.js build/integration/install-alias.integration.test.js",
     "test:smoke": "node --test build/integration/packaged-cli-smoke.test.js",
     "canary:agent-capabilities": "npm run build && node scripts/agent-capabilities-canary.mjs",
     "prepublishOnly": "node scripts/bundle-assets.js && npm run build && node scripts/verify-shebang.cjs"
@@ -52,7 +55,7 @@
     "node": ">=18.0.0"
   },
   "overrides": {
-    "fast-uri": ">=3.1.2",
+    "fast-uri": ">=4.1.1",
     "hono": ">=4.12.31",
     "@hono/node-server": "^1.19.13",
     "ip-address": ">=10.1.1"

package/pipelines/implement-ticket.json

@@ -1,6 +1,6 @@
 {
   "name": "implement-ticket",
-  "description": "Generate plan, execute implementation, commit, open PR, and monitor CI.",
+  "description": "Generate plan, execute implementation, commit, open PR, verify, and monitor CI.",
   "variables": ["ticket_key", "docs_dir"],
   "steps": [
     {
@@ -36,6 +36,11 @@
       "instruction_file": "create-pr.md",
       "description": "Create a pull request with a title derived from the commit subject"
     },
+    {
+      "type": "agent_task",
+      "instruction_file": "verify-plan.md",
+      "description": "Run bounded post-finalization verification and push any correction"
+    },
     {
       "type": "mcp_call",
       "tool": "update_jira_status",

package/build/conductor/supervisor-judgment-python.js

@@ -1,141 +0,0 @@
-/**
- * TypeScript-to-Python judgment adapter (BAPI-396, conductor C4).
- *
- * Lets the Node conductor CLI invoke the approved Python LLM boundary
- * (`src.python.conductor.supervisor_judgment`) WITHOUT any direct model-provider
- * code in TypeScript. The compact judgment request is passed to the Python
- * module over stdin as JSON, the module is spawned with `shell: false` and a
- * list of arguments (never a shell string), and stdout is validated through the
- * shared {@link parseSupervisorJudgmentResponse}.
- *
- * Every boundary failure (missing Python, timeout, non-zero exit, malformed
- * output) is converted to a SANITIZED {@link SupervisorJudgmentError}: the
- * caller ({@link assessSupervisorCandidate}) catches it and degrades to a
- * deterministic assessment. Secrets, stderr text, and stack traces are never
- * placed in the thrown error.
- */
-import { spawn as nodeSpawn } from "node:child_process";
-import path from "node:path";
-import { fileURLToPath } from "node:url";
-import { SupervisorJudgmentError, parseSupervisorJudgmentResponse, } from "./supervisor-judgment.js";
-/** The Python module the adapter invokes via `python -m <module>`. */
-export const SUPERVISOR_JUDGMENT_PYTHON_MODULE = "src.python.conductor.supervisor_judgment";
-function nonEmpty(value) {
-    return typeof value === "string" && value.trim().length > 0;
-}
-/**
- * Resolve the Python executable. Prefers the explicit, safe env override
- * `BAPI_CONDUCTOR_PYTHON`; otherwise falls back to `python3`. No shell string is
- * ever constructed — the returned value is used as `spawn`'s `command` arg.
- */
-export function resolveSupervisorJudgmentCommand(env = process.env) {
-    if (nonEmpty(env.BAPI_CONDUCTOR_PYTHON))
-        return env.BAPI_CONDUCTOR_PYTHON.trim();
-    return "python3";
-}
-/**
- * Resolve the working directory the Python module runs from (the repo root, so
- * `src.python...` imports resolve). Prefers the explicit `BAPI_CONDUCTOR_PYTHON_CWD`
- * override; otherwise derives the repo root relative to this compiled module
- * (`<repo>/mcp_server/build/conductor/<file>.js` -> `<repo>`).
- */
-export function resolveSupervisorJudgmentCwd(env = process.env) {
-    if (nonEmpty(env.BAPI_CONDUCTOR_PYTHON_CWD))
-        return env.BAPI_CONDUCTOR_PYTHON_CWD.trim();
-    const here = fileURLToPath(import.meta.url);
-    // dirname=conductor, ../=build, ../../=mcp_server, ../../../=repo root.
-    return path.resolve(path.dirname(here), "..", "..", "..");
-}
-/** Build the secret-free stdin payload for the Python module. */
-function buildRequestPayload(request, env) {
-    const payload = {
-        run_id: request.run_id,
-        candidate: request.candidate,
-        worker: request.worker,
-    };
-    if (nonEmpty(env.BAPI_CONDUCTOR_REPO_NAME))
-        payload.repo_name = env.BAPI_CONDUCTOR_REPO_NAME.trim();
-    if (nonEmpty(env.BAPI_CONDUCTOR_RUN_ID))
-        payload.session_id = env.BAPI_CONDUCTOR_RUN_ID.trim();
-    return payload;
-}
-/**
- * Spawn the Python judgment module and resolve with the validated response.
- * Rejects with a sanitized {@link SupervisorJudgmentError} on any boundary
- * failure. The request is passed over stdin (keeping payloads out of the process
- * argument list); the timeout is `config.llm_timeout_ms`.
- */
-export function requestPythonSupervisorJudgment(request, config, deps = {}) {
-    const spawnFn = deps.spawn ?? nodeSpawn;
-    const env = deps.env ?? process.env;
-    const command = resolveSupervisorJudgmentCommand(env);
-    const cwd = resolveSupervisorJudgmentCwd(env);
-    return new Promise((resolve, reject) => {
-        let settled = false;
-        let stdout = "";
-        let child;
-        try {
-            child = spawnFn(command, ["-m", SUPERVISOR_JUDGMENT_PYTHON_MODULE], {
-                cwd,
-                shell: false,
-                stdio: ["pipe", "pipe", "pipe"],
-            });
-        }
-        catch {
-            reject(new SupervisorJudgmentError("python judgment process could not be started"));
-            return;
-        }
-        const finish = (fn, value) => {
-            if (settled)
-                return;
-            settled = true;
-            clearTimeout(timer);
-            fn(value);
-        };
-        // NOTE: the timer is intentionally NOT unref'd. It must keep the event loop
-        // alive so the timeout actually fires and rejects when the Python subprocess
-        // hangs (and so the deterministic degraded path is taken). The timer is
-        // always cleared in `finish`, so it never outlives a settled judgment.
-        const timer = setTimeout(() => {
-            try {
-                child.kill("SIGKILL");
-            }
-            catch {
-                /* best-effort */
-            }
-            finish(reject, new SupervisorJudgmentError("python judgment timed out"));
-        }, config.llm_timeout_ms);
-        child.on("error", () => finish(reject, new SupervisorJudgmentError("python judgment process error")));
-        child.stdout?.on("data", (chunk) => {
-            stdout += String(chunk);
-        });
-        child.on("close", (code) => {
-            if (code !== 0) {
-                finish(reject, new SupervisorJudgmentError("python judgment exited non-zero"));
-                return;
-            }
-            try {
-                const parsed = parseSupervisorJudgmentResponse(stdout.trim());
-                finish(resolve, parsed);
-            }
-            catch {
-                finish(reject, new SupervisorJudgmentError("python judgment returned malformed output"));
-            }
-        });
-        try {
-            child.stdin?.write(JSON.stringify(buildRequestPayload(request, env)));
-            child.stdin?.end();
-        }
-        catch {
-            finish(reject, new SupervisorJudgmentError("python judgment stdin write failed"));
-        }
-    });
-}
-/**
- * Build the default injectable judgment client used by the runtime. The returned
- * function matches {@link SupervisorJudgmentClient}; it forwards each request to
- * {@link requestPythonSupervisorJudgment} with the resolved config/deps.
- */
-export function createDefaultSupervisorJudgmentClient(config, deps = {}) {
-    return (request) => requestPythonSupervisorJudgment(request, config, deps);
-}

package/build/conductor/supervisor-judgment.js

@@ -1,215 +0,0 @@
-/**
- * LLM judgment boundary for the conductor supervisor (BAPI-396, conductor C4).
- *
- * The LLM is JUDGMENT-ONLY. It is consulted solely to classify ambiguous stalls
- * and to draft short escalation text — compact snapshot in, strict JSON out. It
- * NEVER executes privileged actions and is NEVER the source of truth: a
- * malformed, disabled, exhausted, or timed-out model call degrades to a
- * deterministic assessment. This module is pure prompt/parse/budget logic with
- * an injectable client and emits NO events.
- */
-/** Raised when a model response cannot be trusted as a valid judgment. */
-export class SupervisorJudgmentError extends Error {
-    constructor(message) {
-        super(message);
-        this.name = "SupervisorJudgmentError";
-    }
-}
-/** Classifications the supervisor accepts from the judgment model. */
-export const ALLOWED_JUDGMENT_CLASSIFICATIONS = new Set([
-    "progressing",
-    "ambiguous",
-    "stuck",
-    "blocked",
-    "unknown",
-]);
-/**
- * Build the judgment-only system prompt. Uses `## HEADERS ##` sections, states
- * the model only classifies stalls / drafts escalation text, prohibits claiming
- * to execute actions, and requires strict JSON output.
- */
-export function buildSupervisorAssessmentSystemPrompt() {
-    return [
-        "## ROLE ##",
-        "You are a read-only watchdog assistant for a multi-agent coding supervisor.",
-        "You classify whether a worker appears stuck and optionally draft a short, human-readable escalation note.",
-        "",
-        "## HARD CONSTRAINTS ##",
-        "- You ONLY classify ambiguous stalls and draft escalation text.",
-        "- You have NO ability to take actions. Never claim to have executed, killed, merged, retried, or fixed anything.",
-        "- You are NEVER the source of truth; deterministic signals override your judgment.",
-        "- Do not request or reveal secrets, tokens, or raw payloads.",
-        "",
-        "## OUTPUT FORMAT ##",
-        "Return STRICT JSON only — no prose, no markdown fences. The object must have exactly these keys:",
-        '  "classification": one of "progressing" | "ambiguous" | "stuck" | "blocked" | "unknown"',
-        '  "confidence": a number between 0 and 1',
-        '  "should_escalate": a boolean',
-        '  "reason": a short machine reason string',
-        '  "draft_escalation_text": a short human-readable string, or null',
-    ].join("\n");
-}
-/**
- * Build the user prompt from a compact judgment request. The dynamic JSON
- * context is wrapped in triple quotes; only compact, secret-free fields are
- * included (the request itself must already be secret-free).
- */
-export function buildSupervisorAssessmentUserPrompt(request) {
-    const compact = {
-        run_id: request.run_id,
-        candidate: {
-            reason: request.candidate.reason,
-            state: request.candidate.state,
-            liveness: request.candidate.liveness,
-            elapsed_ms: request.candidate.elapsed_ms,
-            context: request.candidate.context,
-        },
-        worker: request.worker,
-    };
-    return [
-        "## SUPERVISOR SNAPSHOT ##",
-        "Classify the worker situation below and decide whether it warrants escalation.",
-        "",
-        '"""',
-        JSON.stringify(compact, null, 2),
-        '"""',
-    ].join("\n");
-}
-/** Top-level keys that, if present, mark a response as action-like (rejected). */
-const ACTION_LIKE_KEYS = new Set([
-    "executed",
-    "action",
-    "actions",
-    "command",
-    "commands",
-    "kill",
-    "killed",
-    "merge",
-    "merged",
-    "transition",
-    "deleted",
-    "wrote",
-    "mutated",
-    "ran",
-]);
-/** Phrases in drafted text that assert a privileged action was performed. */
-const ACTION_LIKE_PHRASES = [
-    /\bi (?:have )?(?:killed|merged|executed|ran|deleted|restarted|retried|fixed|committed|pushed)\b/i,
-    /\bhas been (?:killed|merged|executed|restarted|deleted)\b/i,
-    /\bworker (?:killed|terminated|restarted)\b/i,
-];
-/** Parse a JSON value (string or already-parsed object). */
-function asObject(raw) {
-    let value = raw;
-    if (typeof raw === "string") {
-        try {
-            value = JSON.parse(raw);
-        }
-        catch {
-            throw new SupervisorJudgmentError("judgment response is not valid JSON");
-        }
-    }
-    if (value === null || typeof value !== "object" || Array.isArray(value)) {
-        throw new SupervisorJudgmentError("judgment response must be a JSON object");
-    }
-    return value;
-}
-/**
- * Parse + validate a strict-JSON judgment response. Rejects non-objects,
- * malformed JSON, unknown classifications, missing/typed-wrong fields, and any
- * action-like output (forbidden keys or draft text claiming privileged work).
- * Throws {@link SupervisorJudgmentError} on any violation.
- */
-export function parseSupervisorJudgmentResponse(raw) {
-    const obj = asObject(raw);
-    // Reject action-like keys outright.
-    for (const key of Object.keys(obj)) {
-        if (ACTION_LIKE_KEYS.has(key.toLowerCase())) {
-            throw new SupervisorJudgmentError(`judgment response contains forbidden action key "${key}"`);
-        }
-    }
-    const classification = obj.classification;
-    if (typeof classification !== "string" || !ALLOWED_JUDGMENT_CLASSIFICATIONS.has(classification)) {
-        throw new SupervisorJudgmentError("judgment response has an invalid 'classification'");
-    }
-    const confidence = obj.confidence;
-    if (typeof confidence !== "number" || !Number.isFinite(confidence) || confidence < 0 || confidence > 1) {
-        throw new SupervisorJudgmentError("judgment response 'confidence' must be a number in [0,1]");
-    }
-    const shouldEscalate = obj.should_escalate;
-    if (typeof shouldEscalate !== "boolean") {
-        throw new SupervisorJudgmentError("judgment response 'should_escalate' must be a boolean");
-    }
-    const reason = obj.reason;
-    if (typeof reason !== "string" || reason.trim().length === 0) {
-        throw new SupervisorJudgmentError("judgment response 'reason' must be a non-empty string");
-    }
-    let draft = null;
-    const draftRaw = obj.draft_escalation_text;
-    if (draftRaw !== null && draftRaw !== undefined) {
-        if (typeof draftRaw !== "string") {
-            throw new SupervisorJudgmentError("judgment response 'draft_escalation_text' must be a string or null");
-        }
-        for (const pattern of ACTION_LIKE_PHRASES) {
-            if (pattern.test(draftRaw)) {
-                throw new SupervisorJudgmentError("judgment draft text claims a privileged action");
-            }
-        }
-        draft = draftRaw;
-    }
-    return {
-        classification,
-        confidence,
-        should_escalate: shouldEscalate,
-        reason,
-        draft_escalation_text: draft,
-    };
-}
-/** Build the deterministic, degraded-mode assessment for a candidate. */
-function degradedAssessment(candidate, reason) {
-    return {
-        classification: "unknown",
-        confidence: 0,
-        // Degraded mode never SUPPRESSES a surfaced stall: a candidate the
-        // deterministic layer already flagged stays escalated.
-        should_escalate: true,
-        reason: `${candidate.reason}:${reason}`,
-        draft_escalation_text: null,
-        source: "degraded",
-    };
-}
-/**
- * Assess one escalation candidate. Enforces the LLM budget BEFORE calling the
- * injectable client: when the LLM is disabled, the budget is exhausted, or the
- * client times out / errors / returns malformed output, a deterministic
- * degraded-mode assessment is returned. On a successful call `budget.used_calls`
- * is incremented. This function emits NO events.
- */
-export async function assessSupervisorCandidate(request, config, budget, client) {
-    const candidate = request.candidate;
-    if (!config.llm_enabled || !budget.enabled || budget.max_calls <= 0) {
-        return degradedAssessment(candidate, "llm_disabled");
-    }
-    if (budget.used_calls >= budget.max_calls) {
-        return degradedAssessment(candidate, "budget_exhausted");
-    }
-    // Count the attempt against the budget regardless of outcome so a flaky model
-    // cannot be retried unboundedly within a single run.
-    budget.used_calls += 1;
-    let response;
-    try {
-        const raw = await client(request);
-        response = parseSupervisorJudgmentResponse(raw);
-    }
-    catch {
-        return degradedAssessment(candidate, "llm_failed");
-    }
-    return {
-        classification: response.classification,
-        confidence: response.confidence,
-        should_escalate: response.should_escalate,
-        reason: response.reason,
-        draft_escalation_text: response.draft_escalation_text,
-        source: "llm",
-    };
-}