@open-agent-toolkit/cli 0.2.46 → 0.2.48

package/assets/bundle-metadata.json

@@ -1,4 +1,4 @@
 {
   "schemaVersion": 1,
-  "oatVersion": "0.2.46"
+  "oatVersion": "0.2.48"
 }

package/assets/docs/cli-utilities/tool-packs.md

@@ -92,6 +92,25 @@ still fail closed with the owning pack's install or update command. See
 [Writing Skills → Portable sibling reads](../contributing/skills.md#portable-sibling-reads-skills-and-agents)
 for the full contract.
 
+Skills that read a fixed canonical workflows role use a different, local
+`${AGENT_PROVIDER_ROOT}` binding. They try the loaded skill's provider root,
+then `${HOME}/.agents`, then the repository's `.agents` root. The loaded tier
+qualifies only when the exact unsuffixed `agents/<canonical-name>.md` target is
+the same-scope canonical Markdown file or a symlink resolving exactly to it.
+Copies, broken or escaping links, Cursor model/effort variants, and Codex TOML
+roles are misses; resolution continues to the next tier.
+
+That binding only locates canonical source instructions for the one owning
+dependency. It does not select a provider, model, effort, variant, route, or
+native agent. Resolve simultaneous pack dependencies independently. If every
+workflows candidate misses, stop before fallback dispatch and report the
+intended-scope recovery explicitly:
+
+```bash
+oat tools install workflows --scope <user|project>
+oat tools update --pack workflows --scope <user|project>
+```
+
 The packs remain independently installable. If the workflows adapter is
 present without the utility contracts, it fails closed and reports the missing
 dependency instead of inventing a fallback route. Non-project analytical

package/assets/docs/contributing/skills.md

@@ -51,6 +51,37 @@ Probe each candidate for the exact target and use the first match. If none
 exists, name the missing sibling, give an actionable install or update command,
 and stop that workflow branch instead of improvising the sibling's process.
 
+### Skills bind canonical agent reads through a local provider root
+
+When a skill must read one fixed canonical role instruction, bind a local
+`${AGENT_PROVIDER_ROOT}` for that dependency in this order:
+
+1. Derive `${SKILL_DIR}/../..` from the loaded skill path, but admit it only
+   when the exact unsuffixed `agents/<canonical-name>.md` target is the
+   same-scope canonical Markdown file or a symlink whose real path is exactly
+   that file.
+2. Try the user-scope root at `${HOME}/.agents`.
+3. Fall back to the project-scope root at `<repo-root>/.agents`.
+
+An invalid loaded target is a candidate miss, so resolution continues to the
+user and project roots. Regular provider copies, broken or escaping symlinks,
+transformed content, other role names, model/effort/variant-suffixed files, and
+`.codex/agents/*.toml` are never canonical candidates. Claude and Cursor base
+role symlinks can qualify when they resolve exactly to the same-scope canonical
+file; Cursor materialized variants cannot.
+
+`${AGENT_PROVIDER_ROOT}` is local authored-instruction notation, not a global
+environment variable or a provider-native role selector. It has no authority
+over provider, model, effort, variant, route, or dispatch choice. Resolve every
+owning pack independently, even when multiple role dependencies coexist. If
+all candidates miss for a workflows-owned role, stop before fallback launch
+and report both recovery forms for the intended scope:
+
+```bash
+oat tools install workflows --scope <user|project>
+oat tools update --pack workflows --scope <user|project>
+```
+
 ### Materialized agents resolve in two steps
 
 Canonical agents are materialized into provider-specific views — Codex receives
@@ -88,21 +119,26 @@ dependency.
 
 ### The ratchet allows only exact historical baselines
 
-Do not use a bare `.agents/skills/<sibling>/SKILL.md`, `./.agents/...`, or
-`../.agents/...` path for an executable read in any user-default skill or
-agent. The bundled-docs contract test enforces this over the whole
-manifest-derived surface and reports each violation as exact
-`source -> target` evidence.
-
-The only exemption is a pinned historical baseline: non-executable evidence
-such as a dogfood transcript or a fixture README, recorded by exact source
-file, target skill, and target path. There is no wildcard, directory-wide, or
-skill-wide allowance, and no temporary migration allowlist — executable debt is
-fixed at the caller instead of being exempted. The same assertions run against
-the bundled CLI copies of both skills and agents, and the sync contract test
-re-checks the roles it materializes for Codex, so a canonical fix must survive
-bundling and Codex sync as well. No test yet gates the generated `.claude` or
-`.cursor` agent views, so regenerate and spot-check those by hand.
+Do not use a bare `.agents/skills/<sibling>/SKILL.md`,
+`.agents/agents/<canonical-name>.md`, dot-relative equivalent, or repeated-parent
+canonical `agents/<canonical-name>.md` hop for an executable read in any
+user-default skill or agent. The bundled-docs contract test enforces this over
+the whole manifest-derived surface and reports each violation as exact
+`source -> target` evidence. Executable agent findings have a zero baseline, so
+every bare canonical-agent read fails the ratchet. Portable
+`${AGENT_PROVIDER_ROOT}/agents/...` reads and legitimate `.claude` or `.cursor`
+provider-view examples are classified as non-findings.
+
+The pinned historical baseline applies only to skill findings: its six exact
+entries are non-executable evidence such as dogfood transcripts or fixture
+READMEs, recorded by source file, target skill, and target path. There is no
+wildcard, directory-wide, or skill-wide allowance, and no temporary migration
+allowlist — executable debt is fixed at the caller instead of being exempted.
+The same assertions run against the bundled CLI copies of both skills and
+agents, and the sync contract test re-checks the roles it materializes for
+Codex, so a canonical fix must survive bundling and Codex sync as well. No test
+yet gates the generated `.claude` or `.cursor` agent views, so regenerate and
+spot-check those by hand.
 
 ## Contract components
 

package/assets/public-package-versions.json

@@ -1,6 +1,6 @@
 {
-  "cli": "0.2.46",
-  "docs-config": "0.2.46",
-  "docs-theme": "0.2.46",
-  "docs-transforms": "0.2.46"
+  "cli": "0.2.48",
+  "docs-config": "0.2.48",
+  "docs-theme": "0.2.48",
+  "docs-transforms": "0.2.48"
 }

package/assets/skills/oat-project-implement/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: oat-project-implement
-version: 2.3.0
+version: 2.3.1
 description: Use when plan.md is ready for execution. Dispatches one phase implementer per phase, owns independent phase review and bounded fix routing, and supports plan-declared worktree-isolated parallel phases.
 oat_gateable: true
 argument-hint: '[--retry-limit <N>] [--dry-run]'

package/assets/skills/oat-project-implement/references/dispatch-and-dry-run.md

@@ -13,6 +13,32 @@ canonical.
 
 ### Step 0.5: Capability Detection and Tier Selection
 
+**Canonical fallback role instructions:** Before a fresh-child fallback can
+read either workflows-owned role, independently resolve each exact target. For
+`oat-phase-implementer`, bind the first valid root to
+`${IMPLEMENTER_AGENT_PROVIDER_ROOT}`; for `oat-reviewer`, bind the first valid
+root to `${REVIEWER_AGENT_PROVIDER_ROOT}`. For each role independently probe
+candidates in this order: `${SKILL_DIR}/../..`, `${HOME}/.agents`, then
+`<repo-root>/.agents`. Do not reuse one role's resolved root for the other.
+
+Each loaded candidate is valid only when the exact unsuffixed role target is
+the same-scope canonical file or a symlink whose realpath is exactly that
+canonical file. Treat a missing, broken, escaping, copied, transformed,
+suffixed, wrong-role, or otherwise noncanonical candidate as a miss and
+continue to the next candidate. If either required role misses every candidate,
+name that role, stop before the fresh-child fallback, and report
+`oat tools install workflows --scope <user|project>` or
+`oat tools update --pack workflows --scope <user|project>` for the intended
+scope.
+
+These canonical role instructions cannot change the immutable target,
+provider, model, effort, or variant fields selected by the resolver. Native
+dispatch sends the exact resolver-selected variant or `agent_type` first. A
+fresh child is allowed only after an explicit pre-start native role-selection
+rejection before any child starts. Once accepted, continue only through the existing
+handle; terminal results, including `BLOCKED`, never authorize replacement or
+fallback.
+
 **Mandatory target-first order:** Complete Dispatch Policy Preflight and
 resolve the concrete managed target before probing generic agent availability
 or locking Tier 1/Tier 2. A concrete target takes precedence over tier
@@ -22,9 +48,10 @@ native spawn API as native `agent_type`; spawn acceptance plus the constructed
 launcher payload is configured invocation evidence, and independent runtime
 telemetry or agent self-report is not required. Launch a fresh Codex child
 pinned to the resolver target's explicit model, reasoning effort, and
-canonical role instructions from `.agents/agents/oat-phase-implementer.md` or
-`.agents/agents/oat-reviewer.md` only after a native role-selection rejection
-as defined below.
+canonical role instructions from
+`${IMPLEMENTER_AGENT_PROVIDER_ROOT}/agents/oat-phase-implementer.md` or
+`${REVIEWER_AGENT_PROVIDER_ROOT}/agents/oat-reviewer.md` only after a native
+role-selection rejection as defined below.
 If that fresh child cannot be launched, fail closed and block; never
 substitute a base or coordinator role for a managed phase target. Inline
 execution is allowed only with verified equivalent current-host model and

package/assets/skills/oat-project-plan-writing/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: oat-project-plan-writing
-version: 1.2.19
+version: 1.2.20
 description: Use when authoring or mutating plan.md in any OAT workflow. Defines canonical format invariants — stable task IDs, required sections, review table rules, and resume guardrails.
 disable-model-invocation: true
 user-invocable: false
@@ -197,6 +197,25 @@ for an exception. A `## Dispatch Profile` row alone cannot authorize reviewer
 lowering; only a separate reviewed contract may define a bounded lower
 candidate exception.
 
+Before the artifact-review exception can read canonical reviewer instructions,
+resolve the workflows-owned role locally. Probe candidates in this order:
+`${SKILL_DIR}/../..`, `${HOME}/.agents`, then `<repo-root>/.agents`. The loaded
+candidate is valid only when the exact unsuffixed `agents/oat-reviewer.md` is
+the same-scope canonical file or a symlink whose realpath is exactly that
+canonical file. Treat a missing, broken, escaping, copied, transformed,
+suffixed, or otherwise noncanonical candidate as a miss and continue to the
+next candidate. Bind the first valid root to
+`${WORKFLOWS_AGENT_PROVIDER_ROOT}`; never use ambient discovery.
+
+If every candidate misses, name `oat-reviewer`, stop before the artifact-review
+fresh-child fallback dispatch, and report
+`oat tools install workflows --scope <user|project>` or
+`oat tools update --pack workflows --scope <user|project>` for the intended
+scope. Canonical role instructions cannot change the immutable target,
+provider, model, effort, or variant fields selected by the resolver. The exact
+registered native reviewer variant remains first; only an explicit pre-start
+native role-selection rejection before a child starts unlocks the fresh child.
+
 For the exception route, bind every concrete managed reviewer target to the
 actual provider invocation before probing generic reviewer availability or
 selecting execution mechanics. A concrete managed Codex target takes precedence
@@ -207,10 +226,11 @@ over generic tier availability.
   reviewer variant as native `agent_type`. Only a recorded actual pre-start
   native role-selection rejection permits a fresh Codex child with the resolver
   target's explicit model, reasoning effort, and canonical role instructions
-  from `.agents/agents/oat-reviewer.md`. A separately pre-selected CLI route is
-  allowed only when native dispatch cannot express the complete target and no
-  child has started. If another route cannot preserve the target, use only a
-  verified-equivalent inline route or block the review.
+  from `${WORKFLOWS_AGENT_PROVIDER_ROOT}/agents/oat-reviewer.md`. A separately
+  pre-selected CLI route is allowed only when native dispatch cannot express
+  the complete target and no child has started. If another route cannot
+  preserve the target, use only a verified-equivalent inline route or block the
+  review.
 - Claude: require a non-empty `providers.claude.dispatchArgs.model` and put
   that exact value in the actual provider invocation as its `model` argument.
 - Cursor: require a non-empty `providers.cursor.dispatchArgs.variant` and

package/assets/skills/oat-project-review-provide/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: oat-project-review-provide
-version: 1.5.1
+version: 1.5.2
 description: Use when the user explicitly asks to review an OAT project — e.g. "review project", "review the project", "run project review", or confirms a previously offered review. Do NOT auto-invoke on completed work alone. Resolves a project review scope and offers before running.
 disable-model-invocation: false
 user-invocable: true
@@ -624,6 +624,26 @@ Build the "Review Scope" metadata for the reviewer:
 
 ### Step 6: Execute Review (3-Tier Capability Model)
 
+**Canonical reviewer instructions:** Before a direct reviewer-definition read or
+fresh-child fallback, resolve the workflows-owned role locally. Probe candidates
+in this order: `${SKILL_DIR}/../..`, `${HOME}/.agents`, then
+`<repo-root>/.agents`. The loaded candidate is valid only when the exact
+unsuffixed `agents/oat-reviewer.md` is the same-scope canonical file or a
+symlink whose realpath is exactly that canonical file. Treat a missing, broken,
+escaping, copied, transformed, suffixed, or otherwise noncanonical candidate as
+a miss and continue to the next candidate. Bind the first valid root to
+`${WORKFLOWS_AGENT_PROVIDER_ROOT}`; never use ambient discovery.
+
+If every candidate misses, name `oat-reviewer`, stop before launching the
+fresh-child fallback, and report
+`oat tools install workflows --scope <user|project>` or
+`oat tools update --pack workflows --scope <user|project>` for the intended
+scope. Canonical role instructions cannot change the immutable target,
+provider, model, effort, or variant fields selected by the resolver. Native
+dispatch with the exact resolver-selected role remains first. Only an explicit
+pre-start native role-selection rejection before a reviewer starts unlocks a
+fresh child that preserves those fields.
+
 **Step 6.0: Resolve the managed reviewer target**
 
 Before capability-tier selection, resolve the same reviewer contract used by
@@ -669,7 +689,8 @@ Build the actual provider invocation before reporting the target as enforced:
   explicit host rejection of that exact `agent_type` before any reviewer or
   child starts. Only after that pre-start rejection may the workflow launch a
   fresh Codex child with the resolver target's explicit model, reasoning
-  effort, canonical role instructions from `.agents/agents/oat-reviewer.md`,
+  effort, canonical role instructions from
+  `${WORKFLOWS_AGENT_PROVIDER_ROOT}/agents/oat-reviewer.md`,
   and the same Review Scope payload. If neither exact route is available, use
   only a verified-equivalent inline route or block the review.
 - Claude requires a non-empty `providers.claude.dispatchArgs.model`; the actual
@@ -912,7 +933,8 @@ If running inline (Tier 3), execute the review and write artifact.
 6. For final scope: explicitly disposition deferred Medium ledger items (fix now vs accept defer)
 7. Write artifact with file:line references and fix guidance
 
-**Review artifact template:** (see `.agents/agents/oat-reviewer.md` for full format)
+**Review artifact template:** (see
+`${WORKFLOWS_AGENT_PROVIDER_ROOT}/agents/oat-reviewer.md` for full format)
 
 Shared ad-hoc companion reference (non-project mode): the `oat-review-provide`
 review artifact template. Independently probe each required `<name>/SKILL.md`

package/assets/skills/oat-project-review-provide-remote/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: oat-project-review-provide-remote
-version: 1.1.0
+version: 1.1.1
 description: Use when reviewing a GitHub PR opened on another machine for an active OAT project and posting findings back as a single PR review. Resolves the project from the PR diff, reads project artifacts for mode-aware review, and posts via gh api.
 disable-model-invocation: true
 user-invocable: true
@@ -116,7 +116,31 @@ Severity conventions:
 - `medium`: Meaningful but non-blocking quality/maintainability issue.
 - `minor`: Cosmetic/style/documentation issue.
 
-`file` and `line` are both set (inline finding) or both `null` (reviewer-level finding that lands in the top-level body). Checklist + severity model + the `StructuredFindings` schema source of truth: `.agents/agents/oat-reviewer.md` and `design.md` → Data Models → StructuredFindings.
+Before reading the canonical reviewer definition or starting a fresh-child
+fallback, resolve the workflows-owned role locally. Probe candidates in this
+order: `${SKILL_DIR}/../..`, `${HOME}/.agents`, then `<repo-root>/.agents`.
+The loaded candidate is valid only when the exact unsuffixed
+`agents/oat-reviewer.md` is the same-scope canonical file or a symlink whose
+realpath is exactly that canonical file. Treat a missing, broken, escaping,
+copied, transformed, suffixed, or otherwise noncanonical candidate as a miss
+and continue to the next candidate. Bind the first valid root to
+`${WORKFLOWS_AGENT_PROVIDER_ROOT}`; never use ambient discovery.
+
+If every candidate misses, name `oat-reviewer`, stop before starting the
+fresh-child fallback, and report
+`oat tools install workflows --scope <user|project>` or
+`oat tools update --pack workflows --scope <user|project>` for the intended
+scope. Canonical role instructions cannot change the immutable target,
+provider, model, effort, or variant fields selected by the resolver. Native
+dispatch with the exact resolver-selected role remains first. Only an explicit
+pre-start native role-selection rejection before a reviewer starts unlocks a
+fresh child that preserves those fields.
+
+`file` and `line` are both set (inline finding) or both `null` (reviewer-level
+finding that lands in the top-level body). Checklist + severity model + the
+`StructuredFindings` schema source of truth:
+`${WORKFLOWS_AGENT_PROVIDER_ROOT}/agents/oat-reviewer.md` and `design.md` → Data
+Models → StructuredFindings.
 
 ## Process
 
@@ -312,7 +336,10 @@ Detection:
 
 Build the dispatch payload and spawn the reviewer in structured-output mode. This mirrors the tested wrapper at `packages/cli/src/review-remote/reviewer-dispatch.ts` (`buildDispatchPayload` / `dispatchStructuredReview`):
 
-- Set `oat_output_mode: structured` in the dispatch payload (the flag `.agents/agents/oat-reviewer.md` recognizes). In this mode the reviewer returns a `StructuredFindings` object in-memory and writes NO artifact under `reviews/`.
+- Set `oat_output_mode: structured` in the dispatch payload (the flag
+  `${WORKFLOWS_AGENT_PROVIDER_ROOT}/agents/oat-reviewer.md` recognizes). In
+  this mode the reviewer returns a `StructuredFindings` object in-memory and
+  writes NO artifact under `reviews/`.
 - Include the project context (`oat_project`, `oat_review_scope`, `oat_review_head_sha`), the Review Scope metadata block, a pointer to the posted-review-body schema (`design.md` → Data Models → Posted-review-body), and the resolved narrowing range (`<prior_sha>..<HEAD>` or none).
 - If a worktree was resolved in Step 2, include its path so the reviewer reads from the checkout.
 - For Cursor concrete managed dispatch, invoke

package/assets/skills/oat-wave-execute/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: oat-wave-execute
-version: 1.8.2
+version: 1.9.0
 description: Use when executing a wave of external implementation plans as a wrapper OAT project — scaffolding, drift refresh, parallel worktree groups, briefs, gates, merge choreography, and closeout.
 argument-hint: '<wave-id> [plan-names...] (e.g. wave-2 http-listener-before-indexing ...)'
 disable-model-invocation: false
@@ -459,7 +459,8 @@ The mechanical caller constructs an `explainer-kit.run-request/v1` document whos
 required keys are exactly:
 `schemaVersion, recipe, slug, outputRoot, factBase, mode`. Set `recipe` to
 `{ "id": "program-recap", "version": "1" }`; this is an object with exactly
-`id` and `version`. Set `outputRoot` to `.oat/repo/explainers/<slug>/`. Bind the
+`id` and `version`. Set `outputRoot` to
+`.oat/repo/reference/explainers/<slug>/`. Bind the
 synthesized fact-base file through `factBase` with the required keys
 `mode, freshnessPolicy`, set `mode` to `"supplied"`, set `freshnessPolicy` to
 `"live-wins"`, and set `path` to that file.

package/assets/skills/oat-wave-program/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: oat-wave-program
-version: 1.4.0
+version: 1.5.0
 description: Use when decomposing a corpus of external implementation plans into an ordered wave program — coverage inventory, dependency mapping, wave composition, and the durable execution-program artifact that oat-wave-execute consumes and updates.
 argument-hint: '[new|refresh|wave-close <wave-id>] (default: refresh against the current artifact)'
 disable-model-invocation: false
@@ -152,7 +152,8 @@ The mechanical caller constructs an `explainer-kit.run-request/v1` document whos
 required keys are exactly:
 `schemaVersion, recipe, slug, outputRoot, factBase, mode`. Set `recipe` to
 `{ "id": "program-recap", "version": "1" }`; this is an object with exactly
-`id` and `version`. Set `outputRoot` to `.oat/repo/explainers/<slug>/`. Bind the
+`id` and `version`. Set `outputRoot` to
+`.oat/repo/reference/explainers/<slug>/`. Bind the
 synthesized fact-base file through `factBase` with the required keys
 `mode, freshnessPolicy`, set `mode` to `"supplied"`, set `freshnessPolicy` to
 `"live-wins"`, and set `path` to that file.

package/dist/commands/pjm/doctor.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"doctor.d.ts","sourceRoot":"","sources":["../../../src/commands/pjm/doctor.ts"],"names":[],"mappings":"AAUA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,YAAY,CAAC;AAE9C,OAAO,EAAsB,KAAK,WAAW,EAAE,MAAM,YAAY,CAAC;AA4LlE,MAAM,WAAW,gBAAgB;IAC/B,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,QAAQ,CAAC,EAAE,WAAW,CAAC;CACxB;AAED,wBAAsB,kBAAkB,CACtC,QAAQ,EAAE,MAAM,EAChB,OAAO,GAAE,gBAAqB,GAC7B,OAAO,CAAC,WAAW,EAAE,CAAC,CAiVxB"}
+{"version":3,"file":"doctor.d.ts","sourceRoot":"","sources":["../../../src/commands/pjm/doctor.ts"],"names":[],"mappings":"AAUA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,YAAY,CAAC;AAE9C,OAAO,EAAsB,KAAK,WAAW,EAAE,MAAM,YAAY,CAAC;AA4MlE,MAAM,WAAW,gBAAgB;IAC/B,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,QAAQ,CAAC,EAAE,WAAW,CAAC;CACxB;AAED,wBAAsB,kBAAkB,CACtC,QAAQ,EAAE,MAAM,EAChB,OAAO,GAAE,gBAAqB,GAC7B,OAAO,CAAC,WAAW,EAAE,CAAC,CA2UxB"}

package/dist/commands/pjm/doctor.js

@@ -11,9 +11,25 @@ const ALLOWED_TOP_LEVEL_DIRECTORIES = new Set([
     'analysis',
     'reviews',
 ]);
-// A human-facing `README.md` at the repo-reference root is benign, so it is an
-// allowed top-level file alongside the canonical `AGENTS.md` (F5).
-const ALLOWED_TOP_LEVEL_FILES = new Set(['AGENTS.md', 'README.md']);
+// Human/provider-facing companions to the canonical `AGENTS.md` are benign.
+// Pointer correctness remains the responsibility of `oat instructions
+// validate`; PJM doctor only classifies the repository layout.
+const ALLOWED_TOP_LEVEL_FILES = new Set([
+    'AGENTS.md',
+    'CLAUDE.md',
+    'README.md',
+]);
+// `project-observations.md` is the documented append-only project-log ledger.
+// The remaining names have dedicated migration/duplicate checks below, so the
+// loose-reference classifier must not report them a second time.
+const ALLOWED_REFERENCE_FILES = new Set([
+    'AGENTS.md',
+    'CLAUDE.md',
+    'project-observations.md',
+    'decision-record.md',
+    'roadmap.md',
+    'current-state.md',
+]);
 const LEGACY_MONOLITHS = ['reference/decision-record.md'];
 async function pathExists(path) {
     try {
@@ -265,10 +281,7 @@ export async function runPjmDoctorChecks(repoRoot, options = {}) {
     const referenceEntries = await listDirectoryNames(join(repoRoot, 'reference'));
     const looseReferenceFiles = referenceEntries
         .filter((name) => name.endsWith('.md'))
-        .filter((name) => name !== 'AGENTS.md' &&
-        name !== 'decision-record.md' &&
-        name !== 'roadmap.md' &&
-        name !== 'current-state.md')
+        .filter((name) => !ALLOWED_REFERENCE_FILES.has(name))
         .map((name) => `reference/${name}`);
     checks.push({
         name: 'pjm:loose_reference_files',

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@open-agent-toolkit/cli",
-  "version": "0.2.46",
+  "version": "0.2.48",
   "private": false,
   "description": "Open Agent Toolkit CLI",
   "homepage": "https://github.com/voxmedia/open-agent-toolkit/tree/main/packages/cli",
@@ -34,7 +34,7 @@
     "ora": "^9.0.0",
     "yaml": "2.8.2",
     "zod": "^3.25.76",
-    "@open-agent-toolkit/control-plane": "0.2.46"
+    "@open-agent-toolkit/control-plane": "0.2.48"
   },
   "devDependencies": {
     "@types/node": "^22.10.0",