claude-dev-env 2.12.0 → 2.14.0

package/skills/copilot-review/AGENTS.md

@@ -0,0 +1,18 @@
+# copilot-review
+
+Spawns a background subagent that polls the GitHub Copilot reviewer on the current PR, fixes unaddressed findings, and re-requests review each tick until the PR is clean. Triggered by `/copilot-review`, `watch copilot`, `babysit copilot review`, or `keep re-requesting copilot`.
+
+## Purpose
+
+The main session gathers PR context (number, HEAD SHA, owner/repo, branch), spawns a self-terminating background subagent with a fully filled-in prompt, and returns control at once. The subagent loops on a 6-minute `ScheduleWakeup` cadence: fetch Copilot's latest review, TDD-fix any inline findings, push a commit, reply inline, resolve each addressed thread, re-request review. It stops on convergence, a persistent blocker, `TaskStop`, three consecutive ticks with no Copilot review at the current HEAD, or after 20 ticks.
+
+## Key files
+
+| File | Purpose |
+|---|---|
+| `SKILL.md` | Four-step orchestration (opt-out check, gather PR context, spawn subagent, report to user), fix protocol, stop conditions, and ground rules (one commit per tick, honor hooks, preserve draft state, use `copilot-pull-request-reviewer[bot]` with the `[bot]` suffix). |
+| `templates/subagent-prompt.md` | The background watcher prompt Step 3 passes to the subagent word for word, with placeholders for PR number, owner, repo, branch, and HEAD SHA. |
+
+## Environment opt-out
+
+Set `CLAUDE_REVIEWS_DISABLED=copilot` to disable. The skill checks this before spawning anything.

package/skills/copilot-review/CLAUDE.md

@@ -1,18 +1 @@
-# copilot-review
-
-Spawns a background subagent that polls the GitHub Copilot reviewer on the current PR, fixes unaddressed findings, and re-requests review each tick until the PR is clean. Triggered by `/copilot-review`, `watch copilot`, `babysit copilot review`, or `keep re-requesting copilot`.
-
-## Purpose
-
-The main session gathers PR context (number, HEAD SHA, owner/repo, branch), spawns a self-terminating background subagent with a fully filled-in prompt, and returns control at once. The subagent loops on a 6-minute `ScheduleWakeup` cadence: fetch Copilot's latest review, TDD-fix any inline findings, push a commit, reply inline, resolve each addressed thread, re-request review. It stops on convergence, a persistent blocker, `TaskStop`, three consecutive ticks with no Copilot review at the current HEAD, or after 20 ticks.
-
-## Key files
-
-| File | Purpose |
-|---|---|
-| `SKILL.md` | Four-step orchestration (opt-out check, gather PR context, spawn subagent, report to user), fix protocol, stop conditions, and ground rules (one commit per tick, honor hooks, preserve draft state, use `copilot-pull-request-reviewer[bot]` with the `[bot]` suffix). |
-| `templates/subagent-prompt.md` | The background watcher prompt Step 3 passes to the subagent word for word, with placeholders for PR number, owner, repo, branch, and HEAD SHA. |
-
-## Environment opt-out
-
-Set `CLAUDE_REVIEWS_DISABLED=copilot` to disable. The skill checks this before spawning anything.
+@AGENTS.md

package/skills/descriptions/SKILL.md

@@ -0,0 +1,51 @@
+---
+name: descriptions
+description: >-
+  Draft or revise a pull-request description. Use when a user asks for a "PR
+  description", "pull request summary", or "change description".
+---
+
+# Pull Request Description Guide
+
+## When to Use This Guide
+
+Use this guide to prepare or revise the written context for a pull request. A
+description explains the change to reviewers and future readers; it records the
+intent and evidence that source code alone cannot express.
+
+## Required Content
+
+Start with a short, specific summary line. Follow it with the information needed
+to understand the change:
+
+- The problem, motivation, and affected users or systems.
+- The implemented approach and meaningful design decisions.
+- Verification performed, including test commands and observed results.
+- Risks, migrations, compatibility effects, and rollback or recovery context.
+- Related issues, design records, and scoped follow-up work.
+
+Use durable prose for every essential fact when linked material has restricted
+access or a limited retention period.
+
+## Description Shape
+
+```text
+Summarize the delivered change.
+
+Explain the problem and the approach.
+
+Verification: <commands and results>
+Risk: <migration, compatibility, or recovery context>
+```
+
+Keep tags short and use them according to the repository's established format.
+
+## Reviewing a Description
+
+Compare the description with the changed files and the review findings. Keep the
+summary, scope, verification, and risk information aligned with the pull
+request's current contents.
+
+Use [reviews](../reviews/SKILL.md#review-workflow) for substantive code
+assessment and [small-cl](../small-cl/SKILL.md#what-a-focused-pull-request-contains)
+to establish a focused scope.

package/skills/emergencies/SKILL.md

@@ -0,0 +1,42 @@
+---
+name: emergencies
+description: >-
+  Classify an urgent production change and guide its focused emergency review.
+  Use when a user asks about an "emergency hotfix", "production incident", or "urgent security fix".
+---
+
+# Emergency Review Guide
+
+## Emergency Classification
+
+Use emergency handling for a narrow change that mitigates an active, severe
+consequence:
+
+- A production failure materially affecting users or a critical service.
+- An active security or legal exposure requiring prompt mitigation.
+- A contractual, hardware, or market deadline where a missed delivery causes a
+  severe external consequence.
+
+Record the incident, affected system, immediate consequence, and the proposed
+mitigation before assessing the change.
+
+## Standard Priority
+
+Use the [review guide](../reviews/SKILL.md#review-workflow) for feature
+schedules, end-of-day timing, ordinary delivery targets, and build remediation.
+Use a focused rollback or remediation change when it restores service safely.
+
+## While the Incident Is Active
+
+Keep the change as small as the mitigation permits. Evaluate whether it resolves
+the incident, introduces a new immediate risk, and has sufficient validation for
+the affected path.
+
+State the incident context, verification performed, observed result, and any
+remaining recovery work in the pull request. Use the [comment guide](../comments/SKILL.md#writing-useful-review-comments)
+to communicate required mitigation work clearly.
+
+## Follow-up Assessment
+
+After service is stable, use the [review guide](../reviews/SKILL.md#evaluation-criteria)
+for the broader maintainability, testing, documentation, and prevention work.

package/skills/everything-search/AGENTS.md

@@ -0,0 +1,17 @@
+# everything-search
+
+Fast file search on Windows using the Everything (voidtools) `es.exe` command-line tool. Triggered by `find files`, `search for files`, `locate files`, or any request to use Everything.
+
+## Purpose
+
+Searches files by extension, name, date modified, size, or path against Everything's real-time index. Returns results instantly regardless of drive size. The skill covers the correct `es.exe` path, search operator syntax, multi-extension patterns, output options, and the junction/symlink note (Everything indexes real NTFS paths only — not junctions or mapped drives).
+
+## Key file
+
+| File | Purpose |
+|---|---|
+| `SKILL.md` | Command syntax for WSL and Windows paths, all search operators (`ext:`, `dm:`, `size:`, wildcards, OR/AND/NOT), output flags (`-sort`, `-n`, `-size`), worked examples, and a junction/drive-mapping lookup table. |
+
+## When to use
+
+Use for file-system searches by name, extension, size, or date. For content searches use Grep. If a path is a junction or mapped drive and Everything returns nothing, translate to the real NTFS path or fall back to the Glob tool.

package/skills/everything-search/CLAUDE.md

@@ -1,17 +1 @@
-# everything-search
-
-Fast file search on Windows using the Everything (voidtools) `es.exe` command-line tool. Triggered by `find files`, `search for files`, `locate files`, or any request to use Everything.
-
-## Purpose
-
-Searches files by extension, name, date modified, size, or path against Everything's real-time index. Returns results instantly regardless of drive size. The skill covers the correct `es.exe` path, search operator syntax, multi-extension patterns, output options, and the junction/symlink note (Everything indexes real NTFS paths only — not junctions or mapped drives).
-
-## Key file
-
-| File | Purpose |
-|---|---|
-| `SKILL.md` | Command syntax for WSL and Windows paths, all search operators (`ext:`, `dm:`, `size:`, wildcards, OR/AND/NOT), output flags (`-sort`, `-n`, `-size`), worked examples, and a junction/drive-mapping lookup table. |
-
-## When to use
-
-Use for file-system searches by name, extension, size, or date. For content searches use Grep. If a path is a junction or mapped drive and Everything returns nothing, translate to the real NTFS path or fall back to the Glob tool.
+@AGENTS.md

package/skills/fresh-branch/AGENTS.md

@@ -0,0 +1,14 @@
+# fresh-branch
+
+Creates a new branch from fresh-fetched `origin/main` inside an isolated worktree under a configured root. Default: `<repo-root>/.claude/worktrees/<agent>/<branch-name>`. Optional absolute `--worktree-root` relocates the root (agent and branch still nest under it). Relative roots and path escape fail closed. Does not push, open a PR, or run `checkout -b` in the caller tree.
+
+## Key files
+
+| File | Purpose |
+|---|---|
+| `SKILL.md` | Phases, checklist, configured-root contract, execute-vs-read for the CLI, gotchas |
+| `scripts/create_fresh_branch.py` | Deterministic CLI: resolve configured root (fail closed before fetch), fetch base, `git worktree add -b`, JSON stdout |
+| `scripts/fresh_branch_git_commands.py` | Git command helpers: fetch, ref checks, `git worktree add -b --no-track` |
+| `scripts/test_create_fresh_branch.py` | Behavioral tests with temporary git repos (default root, explicit root, escape, collisions) |
+| `scripts/test_fresh_branch_git_commands.py` | Behavioral tests for the git command helpers |
+| `scripts/fresh_branch_scripts_constants/` | Constants package (`fresh_branch_cli_constants`) for CLI flags and error strings |

package/skills/fresh-branch/CLAUDE.md

@@ -1,14 +1 @@
-# fresh-branch
-
-Creates a new branch from fresh-fetched `origin/main` inside an isolated worktree under a configured root. Default: `<repo-root>/.claude/worktrees/<agent>/<branch-name>`. Optional absolute `--worktree-root` relocates the root (agent and branch still nest under it). Relative roots and path escape fail closed. Does not push, open a PR, or run `checkout -b` in the caller tree.
-
-## Key files
-
-| File | Purpose |
-|---|---|
-| `SKILL.md` | Phases, checklist, configured-root contract, execute-vs-read for the CLI, gotchas |
-| `scripts/create_fresh_branch.py` | Deterministic CLI: resolve configured root (fail closed before fetch), fetch base, `git worktree add -b`, JSON stdout |
-| `scripts/fresh_branch_git_commands.py` | Git command helpers: fetch, ref checks, `git worktree add -b --no-track` |
-| `scripts/test_create_fresh_branch.py` | Behavioral tests with temporary git repos (default root, explicit root, escape, collisions) |
-| `scripts/test_fresh_branch_git_commands.py` | Behavioral tests for the git command helpers |
-| `scripts/fresh_branch_scripts_constants/` | Constants package (`fresh_branch_cli_constants`) for CLI flags and error strings |
+@AGENTS.md

package/skills/grok-spawn/AGENTS.md

@@ -0,0 +1,28 @@
+# grok-spawn
+
+**Trigger:** `/grok-spawn`, "spawn grok workers", "grok worker fleet", "run this with grok workers", "headless grok batch".
+
+Orchestrator playbook for fleets of headless grok CLI workers. Composes
+`grok_worker_preflight.py` and `spawn_grok_batch.py` by name. Holds no spawn
+logic of its own. The calling session verifies results and owns every commit,
+push, and GitHub post.
+
+## Subdirectories
+
+| Directory | Role |
+|---|---|
+| `reference/` | Prompt-part templates and tool-profile flag docs the hub cites on demand. |
+
+## Key files
+
+| File | Role |
+|---|---|
+| `SKILL.md` | Entry protocol: when to use, gotchas, preflight → batch spec → launch → collect → verify. |
+| `reference/worker-briefs.md` | Readonly investigation brief, build brief (edits and tests only), report contract. |
+| `reference/flag-profiles.md` | Readonly vs build flag sets, shared flags, leader-socket and stagger rules, `--agent` note. |
+
+## Entry point
+
+No skill-local scripts. Runtime lives in `$HOME/.claude/scripts/`
+(`grok_worker_preflight.py`, `spawn_grok_batch.py`, `grok_headless_runner.py`)
+after install. Constants live in `scripts/dev_env_scripts_constants/`.

package/skills/grok-spawn/CLAUDE.md

@@ -1,28 +1 @@
-# grok-spawn
-
-**Trigger:** `/grok-spawn`, "spawn grok workers", "grok worker fleet", "run this with grok workers", "headless grok batch".
-
-Orchestrator playbook for fleets of headless grok CLI workers. Composes
-`grok_worker_preflight.py` and `spawn_grok_batch.py` by name. Holds no spawn
-logic of its own. The calling session verifies results and owns every commit,
-push, and GitHub post.
-
-## Subdirectories
-
-| Directory | Role |
-|---|---|
-| `reference/` | Prompt-part templates and tool-profile flag docs the hub cites on demand. |
-
-## Key files
-
-| File | Role |
-|---|---|
-| `SKILL.md` | Entry protocol: when to use, gotchas, preflight → batch spec → launch → collect → verify. |
-| `reference/worker-briefs.md` | Readonly investigation brief, build brief (edits and tests only), report contract. |
-| `reference/flag-profiles.md` | Readonly vs build flag sets, shared flags, leader-socket and stagger rules, `--agent` note. |
-
-## Entry point
-
-No skill-local scripts. Runtime lives in `$HOME/.claude/scripts/`
-(`grok_worker_preflight.py`, `spawn_grok_batch.py`, `grok_headless_runner.py`)
-after install. Constants live in `scripts/dev_env_scripts_constants/`.
+@AGENTS.md

package/skills/orchestrator/SKILL.md

@@ -232,7 +232,7 @@ workflow resume is available.
 | Work | Agent type | Model |
 |---|---|---|
 | Feature, bug, and refactor coding | `clean-coder` | `sonnet` on a Claude host; the sonnet-equivalent id the worker-model resolver prints on a third-party host |
-| Verification passes | `code-verifier` | `sonnet` on a Claude host; the sonnet-equivalent id the worker-model resolver prints on a third-party host |
+| Review and verification | `code-quality-agent` | `sonnet` on a Claude host; the sonnet-equivalent id the worker-model resolver prints on a third-party host |
 | Script runs, GitHub posting, and backfill driving | `general-purpose` runner | `sonnet` on a Claude host; the sonnet-equivalent id the worker-model resolver prints on a third-party host |
 | PR descriptions | `pr-description-writer` | `haiku`, with file-list grounding check |
 | Fan-out searches and checklist verification reads | `Explore` | `haiku`; use `sonnet` when judgment-heavy |
@@ -265,8 +265,7 @@ Routing rules:
   system, with no second one.
 - Resume a warm workflow agent before creating a new workflow run when
   the warm agent holds the relevant context.
-- `clean-coder` owns code edits. `code-verifier` owns verification. The
-  same workflow agent never grades work it wrote.
+- Review and verification workflows apply the [review guide](../reviews/SKILL.md#review-workflow).
 - PR-description workflows include the actual changed-file list in the
   prompt and verify the final body against that file list before posting
   or returning it.
@@ -363,4 +362,3 @@ no open work remains, run `set --status done` before any re-arm attempt.
 ## File-backed run ledger
 
 When host task tools are absent, reconcile delegated work through `scripts/grok_run_ledger.py` under the run-state directory (stable task ids, one live owner, unique advisor sessions, dependency blocking, snapshot-drift reopening).
-

package/skills/plan-to-pr/SKILL.md

@@ -57,8 +57,7 @@ Refuse the first matching case and stop:
 - A requested change spans files outside the packet's allowed file set.
 - A task contains more than one deliverable, acceptance check, or commit.
 - A required task tool, verifier, model tier, advisor, review path, or gate is unavailable.
-- The user asks to bypass verification, `verified_commit_gate`, independent
-  review, final validation, or a publication gate.
+- The user asks to bypass verification, review under the [review guide](../reviews/SKILL.md#review-workflow), final validation, or a publication gate.
 
 Return exactly: `Plan-to-PR blocked: <missing input or capability>.`
 
@@ -105,8 +104,8 @@ For each seeded task:
 1. The orchestrator issues the standalone task ticket.
 2. One worker implements one deliverable within the ticket's allowed files.
 3. The worker runs the one acceptance check and reports exact output and blockers.
-4. A fresh verifier checks the task diff, baseline, named gates, and ticket-to-diff
-   scope. It must pass `verified_commit_gate` for the exact surface.
+4. Review and verify the task diff, baseline, named checks, and ticket-to-diff
+   scope under the [review guide](../reviews/SKILL.md#review-workflow).
 5. The orchestrator creates exactly one commit and records its hash.
 
 No task borrows files, acceptance checks, or commit history from another task.
@@ -114,7 +113,7 @@ Unrelated changes remain untouched and uncommitted.
 
 ## Review and repair
 
-Before every commit, require fresh verification and `verified_commit_gate`.
+Before every commit, record fresh verification and review under the [review guide](../reviews/SKILL.md#review-workflow).
 After the commit, a separate fast low-effort Luna review worker runs native
 findings-only correctness review at `/e-code-review low`. The review returns
 findings only and has no repair flag.
@@ -128,8 +127,8 @@ gate after cleanup.
 ## Final validation and publication
 
 The Luna max final validator maps every commit to one packet task and checks
-cumulative behavior, the allowed-file ledger, acceptance checks, verifier
-output, `verified_commit_gate` evidence, and post-commit review records. Any
+cumulative behavior, the allowed-file ledger, acceptance checks, verification
+output, and post-commit review records. Any
 unmapped commit, missing record, failed gate, or unresolved finding blocks
 publication.
 

package/skills/plan-to-pr/reference/final-validation-tasks.md

@@ -7,7 +7,7 @@ PASS, FAIL plus file and line evidence, or N/A plus a reason.
 2. Confirm every task record has required identity, files, acceptance, and commit fields.
 3. Confirm every committed task has native findings-only review and separate repair records.
 4. Confirm every repair has acceptance output, fresh exact-surface verification, and an amended commit before the next review.
-5. Confirm all review records are clean and all verifier records pass.
+5. Confirm all review records are clean and all verification records are present.
 6. Confirm the complete allowed-file ledger contains no extra changed files.
 7. Confirm final validation and workflow self-audit evidence are present before release.
 8. Confirm the finalized PR branch receives Luna max `/e-simplify` cleanup with

package/skills/plan-to-pr/reference/process-inventory.md

@@ -11,10 +11,10 @@ judgment remains in the hub or companion reference.
 | Seed implementation work after packet approval | deterministic | `task-seed:reference/task-seeds.md` | Host task IDs, packet path, and evidence | task-tool |
 | Record task fields | deterministic | `reference/run-record.schema.json` and `scripts/validate_protocol.py` | Validator output and exit code | `scripts/test_validate_protocol.py` |
 | Implement one deliverable | judgment | `SKILL.md` | Worker report and diff | N/A: implementation choice |
-| Verify exact surface | deterministic | `task-seed:reference/task-seeds.md` | Verifier output and `verified_commit_gate` | task-tool |
+| Verify exact surface | deterministic | `task-seed:reference/task-seeds.md` | Verification output and [review guide](../../reviews/SKILL.md#review-workflow) | task-tool |
 | Review committed task | deterministic | `reference/review-loop.md` | Findings-only native review record | `test_task_ticket_contract.py` |
 | Repair confirmed findings | deterministic | `reference/review-loop.md` | Separate repair record and confirmed findings | `test_task_ticket_contract.py` |
-| Decide whether a finding is confirmed | judgment | `reference/review-loop.md` | Repair record names the confirmed finding | N/A: review judgment |
+| Decide whether a finding is confirmed | judgment | [review guide](../../reviews/SKILL.md#review-judgment) | Repair record names the confirmed finding | N/A: review judgment |
 | Repeat until clean | borderline | `reference/review-loop.md` | Amended commit and clean review record | `test_task_ticket_contract.py` |
 | Validate cumulative release | deterministic | `task-seed:reference/final-validation-tasks.md` | Final validator task evidence | task-tool |
 | Self-audit the workflow | deterministic | `task-seed:reference/self-audit-tasks.md` | Self-audit task evidence | task-tool |

package/skills/plan-to-pr/reference/review-loop.md

@@ -6,7 +6,7 @@ Run this loop for every committed task. Store each pass in the task record.
 2. It invokes the native findings-only `/e-code-review low` correctness
    capability; the review has no repair flag.
 3. Record resolved model, effort, command, findings, repair status, and surface
-   hash. Missing native review or verifier capability fails closed.
+   hash. Apply the [review guide](../../reviews/SKILL.md#review-workflow) to every finding.
 4. A separate fast low-effort Luna repair worker applies only confirmed findings.
 5. Rerun the task acceptance check and fresh exact-surface verification.
 6. Amend the task commit and repeat the native review until clean.
@@ -14,7 +14,7 @@ Run this loop for every committed task. Store each pass in the task record.
 ## Accepted records
 
 Clean means the native review reports no findings for the exact committed
-surface. A repaired pass is not accepted until acceptance and fresh verifier
+surface. A repaired pass is not accepted until acceptance and fresh verification
 output are recorded against the amended commit.
 
 ## Example

package/skills/plan-to-pr/reference/run-record.schema.json

@@ -22,6 +22,6 @@
     "finding": {"type": "object", "additionalProperties": false, "required": ["finding", "disposition"], "properties": {"finding": {"type": "string", "minLength": 1}, "disposition": {"enum": ["confirmed", "dismissed"]}}},
     "review_record": {"type": "object", "additionalProperties": false, "required": ["resolved_model", "effort", "command", "findings", "repair_status", "surface_hash", "findings_only", "has_repair_flag"], "properties": {"resolved_model": {"type": "string", "minLength": 1}, "effort": {"type": "string", "minLength": 1}, "command": {"type": "string", "minLength": 1}, "findings": {"type": "array", "items": {"$ref": "#/$defs/finding"}}, "repair_status": {"type": "string", "minLength": 1}, "surface_hash": {"type": "string", "minLength": 1}, "findings_only": {"const": true}, "has_repair_flag": {"const": false}}},
     "repair_record": {"type": "object", "additionalProperties": false, "required": ["resolved_model", "effort", "confirmed_findings", "repair_status", "surface_hash"], "properties": {"resolved_model": {"type": "string", "minLength": 1}, "effort": {"type": "string", "minLength": 1}, "confirmed_findings": {"type": "array", "items": {"type": "string", "minLength": 1}}, "repair_status": {"type": "string", "minLength": 1}, "surface_hash": {"type": "string", "minLength": 1}}},
-    "verification_record": {"type": "object", "additionalProperties": false, "required": ["acceptance_output", "verifier_output", "verified_commit_gate", "surface_hash"], "properties": {"acceptance_output": {"type": "string", "minLength": 1}, "verifier_output": {"type": "string", "minLength": 1}, "verified_commit_gate": {"type": "string", "minLength": 1}, "surface_hash": {"type": "string", "minLength": 1}}}
+    "verification_record": {"type": "object", "additionalProperties": false, "required": ["acceptance_output", "verification_output", "surface_hash"], "properties": {"acceptance_output": {"type": "string", "minLength": 1}, "verification_output": {"type": "string", "minLength": 1}, "surface_hash": {"type": "string", "minLength": 1}}}
   }
 }

package/skills/plan-to-pr/reference/task-seeds.md

@@ -7,7 +7,7 @@ equivalent before implementation. Mark it complete only with evidence.
 2. Capture the baseline and create one task record from the run-record schema.
 3. Implement the single deliverable within the allowed file set.
 4. Run the acceptance check and record its exact output.
-5. Run fresh exact-surface verification and `verified_commit_gate`.
+5. Run fresh exact-surface verification and apply the [review guide](../../reviews/SKILL.md#review-workflow).
 6. Create exactly one task commit and record its identity.
 7. Run the separate native review and record findings-only output.
 8. Run separate Luna repair only for confirmed findings; rerun acceptance and

package/skills/plan-to-pr/reference/task-ticket.md

@@ -17,13 +17,12 @@ commits.
 | Worker route | The model role and selected route from `reference/model-routing.md`; planner and final-validator records include Luna max plus Sol xhigh advisor evidence, and orchestrator records the max route |
 | Commit record | Exactly one commit hash for this ticket |
 | Review record | Separate fast low-effort Luna review output that invokes native `/e-code-review low` and returns findings only, followed by separate fast low-effort Luna repair output, including resolved model, effort, command, findings, repair status, and surface hash |
-| Verification record | Fresh verifier output and `verified_commit_gate` result for the exact surface |
+| Verification record | Fresh verification output for the exact surface and review under the [review guide](../../reviews/SKILL.md#review-workflow) |
 
 ## Execution contract
 
 The worker changes only the allowed files, runs the acceptance check, and reports
-the exact output. A fresh verifier checks the ticket against the diff before one
-commit is created. The commit hash is recorded before the separate review pass.
+the exact output. Review and verification follow the [review guide](../../reviews/SKILL.md#review-workflow) before one commit is created. The commit hash is recorded before the separate review pass.
 
 ## Review and completion contract
 
@@ -33,6 +32,6 @@ repair flag. A separate fast low-effort Luna repair worker applies only confirme
 the resolved model, effort, command, findings, repair status, and surface hash.
 Confirmed repairs require the acceptance check and fresh exact-surface verification
 again; amend the task commit and repeat the native review until clean. Missing
-native review or required verifier capability fails closed. The
+native review fails closed. The
 final validator maps the commit to this ticket and rejects missing or extra
 records.

package/skills/plan-to-pr/scripts/config/constants.py

@@ -174,8 +174,7 @@ ALL_REPAIR_FIELDS: tuple[str, ...] = (
 )
 ALL_VERIFICATION_FIELDS: tuple[str, ...] = (
     "acceptance_output",
-    "verifier_output",
-    "verified_commit_gate",
+    "verification_output",
     "surface_hash",
 )
 ALL_REVIEW_TEXT_FIELDS: tuple[str, ...] = (

package/skills/plan-to-pr/scripts/test_validate_protocol.py

@@ -20,7 +20,7 @@ validate_record = VALIDATE_PROTOCOL_MODULE.validate_record
 
 
 def valid_record() -> dict[str, object]:
-    verification = {"acceptance_output": "passed: pytest", "verifier_output": "passed: clean", "verified_commit_gate": "passed: exact surface", "surface_hash": "a" * 64}
+    verification = {"acceptance_output": "passed: pytest", "verification_output": "passed: clean", "surface_hash": "a" * 64}
     return {
         "task_identity": "task-3", "deliverable": "validator", "allowed_files": ["validator.py"],
         "acceptance_check": "pytest", "baseline": "clean", "worker_route": "implementation worker; effort=low", "commit": "a1b2c3d",

package/skills/plan-to-pr/test_skill_contract.py

@@ -90,7 +90,7 @@ def test_skill_contract_enforces_task_commit_and_review_order() -> None:
     skill_text = read_skill_text()
 
     assert EXPECTED_TASK_PROTOCOL_HEADING in skill_text
-    assert "fresh verification and `verified_commit_gate`" in skill_text
+    assert "fresh verification and review under the [review guide]" in skill_text
     assert (
         "native\nfindings-only correctness review at `/e-code-review low`" in skill_text
     )
@@ -163,7 +163,7 @@ def test_skill_contract_references_future_fixed_artifacts_without_copying_tables
 
 def test_skill_contract_companion_reference_paths_exist() -> None:
     skill_text = read_skill_text()
-    local_reference_paths = re.findall(r"\]\(([^)]+)\)", skill_text)
+    local_reference_paths = [each_path.split("#", maxsplit=1)[0] for each_path in re.findall(r"\]\(([^)]+)\)", skill_text)]
     repository_paths = [
         each_path
         for each_path in local_reference_paths

package/skills/pr-converge/AGENTS.md

@@ -0,0 +1,32 @@
+# pr-converge skill
+
+Drives a draft PR to convergence by driving the internal passes to clean first — a code-review pass and a bugteam audit — then running Cursor Bugbot and Copilot as terminal confirmation gates, applying TDD fixes, posting inline replies, and re-triggering the external reviewers each tick until all are clean on the same HEAD and the PR is mergeable.
+
+**Trigger:** `/pr-converge`, "drive PR to convergence", "loop bugbot and bugteam", "babysit bugbot and bugteam", "until both are clean", "converge this PR".
+
+## Key files
+
+| File | Purpose |
+|---|---|
+| `SKILL.md` | Full tick workflow, pre-flight checks, state schema, budget-aware tick boundaries, stop conditions |
+| `pr_converge_skill_constants/constants.py` | Runtime and API constants (bot logins, review states, GH API path templates, regex patterns) |
+
+## Subdirectories
+
+| Directory | Role |
+|---|---|
+| `pr_converge_skill_constants/` | Importable constants module shared by skill scripts |
+| `reference/` | Per-tick steps, convergence gates, fix protocol, obstacle runbooks, state schema, stop conditions, examples |
+| `scripts/` | Python helpers (bugbot check, convergence check, Copilot review fetcher, fix-reply poster, reflow tool) and their tests |
+| `workflows/` | ScheduleWakeup loop pacing specification (`pacer=schedule_wakeup`) |
+
+## Conventions
+
+- Pre-flight selects a pacer via `select_converge_pacer.py`: `schedule_wakeup` or `portable`. Missing `ScheduleWakeup` selects portable — it does not abort the skill.
+- On `pacer=schedule_wakeup`, each invocation runs one tick and the next advances via `ScheduleWakeup`. On `pacer=portable`, the session runs ticks continuously (continue or in-session poll) until convergence, a stop condition, or a budget handoff.
+- Loop state persists to `$CLAUDE_JOB_DIR/pr-converge-state.json` between ticks.
+- Isolation uses `EnterWorktree` when present; otherwise the portable driver git worktree path, then `preflight_worktree.py`.
+- All findings and PR reports state verified facts only — no hedging language.
+- The GitHub MCP (`pull_request_read`, `pull_request_review_write`) is the primary path for PR inspection; `gh api` is the fallback.
+- Three step-scoped agents (`fix_executor`, `thread_sweep`, `copilot_watch`) persist across ticks via the `persistent_agents` map in loop state; each tick resumes them with `SendMessage` and spawns a fresh named agent when a stored id is dead.
+- The tick delegates shared mechanics to three sibling sub-skills: `reviewer-gates` (opt-out, Copilot quota, Bugbot trigger), `pr-fix-protocol` (fix sequence + unresolved-thread sweep), and `pr-loop-lifecycle` (run open/close). Target resolution happens inline in `reference/per-tick.md` Steps 1 and 1.5. Audit posting follows the `_shared/pr-loop/post-audit-thread-contract.md` contract, run through `_shared/pr-loop/scripts/post_audit_thread.py`.

package/skills/pr-converge/CLAUDE.md

@@ -1,32 +1 @@
-# pr-converge skill
-
-Drives a draft PR to convergence by driving the internal passes to clean first — a code-review pass and a bugteam audit — then running Cursor Bugbot and Copilot as terminal confirmation gates, applying TDD fixes, posting inline replies, and re-triggering the external reviewers each tick until all are clean on the same HEAD and the PR is mergeable.
-
-**Trigger:** `/pr-converge`, "drive PR to convergence", "loop bugbot and bugteam", "babysit bugbot and bugteam", "until both are clean", "converge this PR".
-
-## Key files
-
-| File | Purpose |
-|---|---|
-| `SKILL.md` | Full tick workflow, pre-flight checks, state schema, budget-aware tick boundaries, stop conditions |
-| `pr_converge_skill_constants/constants.py` | Runtime and API constants (bot logins, review states, GH API path templates, regex patterns) |
-
-## Subdirectories
-
-| Directory | Role |
-|---|---|
-| `pr_converge_skill_constants/` | Importable constants module shared by skill scripts |
-| `reference/` | Per-tick steps, convergence gates, fix protocol, obstacle runbooks, state schema, stop conditions, examples |
-| `scripts/` | Python helpers (bugbot check, convergence check, Copilot review fetcher, fix-reply poster, reflow tool) and their tests |
-| `workflows/` | ScheduleWakeup loop pacing specification (`pacer=schedule_wakeup`) |
-
-## Conventions
-
-- Pre-flight selects a pacer via `select_converge_pacer.py`: `schedule_wakeup` or `portable`. Missing `ScheduleWakeup` selects portable — it does not abort the skill.
-- On `pacer=schedule_wakeup`, each invocation runs one tick and the next advances via `ScheduleWakeup`. On `pacer=portable`, the session runs ticks continuously (continue or in-session poll) until convergence, a stop condition, or a budget handoff.
-- Loop state persists to `$CLAUDE_JOB_DIR/pr-converge-state.json` between ticks.
-- Isolation uses `EnterWorktree` when present; otherwise the portable driver git worktree path, then `preflight_worktree.py`.
-- All findings and PR reports state verified facts only — no hedging language.
-- The GitHub MCP (`pull_request_read`, `pull_request_review_write`) is the primary path for PR inspection; `gh api` is the fallback.
-- Three step-scoped agents (`fix_executor`, `thread_sweep`, `copilot_watch`) persist across ticks via the `persistent_agents` map in loop state; each tick resumes them with `SendMessage` and spawns a fresh named agent when a stored id is dead.
-- The tick delegates shared mechanics to three sibling sub-skills: `reviewer-gates` (opt-out, Copilot quota, Bugbot trigger), `pr-fix-protocol` (fix sequence + unresolved-thread sweep), and `pr-loop-lifecycle` (run open/close). Target resolution happens inline in `reference/per-tick.md` Steps 1 and 1.5. Audit posting follows the `_shared/pr-loop/post-audit-thread-contract.md` contract, run through `_shared/pr-loop/scripts/post_audit_thread.py`.
+@AGENTS.md

package/skills/pr-converge/SKILL.md

@@ -168,9 +168,8 @@ when `$CLAUDE_JOB_DIR` is gone.
 
 Fields: `phase`, `tick_count`, `bugbot_clean_at`, `code_review_clean_at`,
 `bugteam_clean_at`, `copilot_clean_at`, `merge_state_status`, `current_head`,
-`bugbot_acknowledged_at`, `bugbot_down`, `copilot_down`,
-`bugteam_skill_invoked_at_head`, `bugteam_skill_invoked_at_tick`,
-`agents_session_id`, `persistent_agents`.
+`bugbot_acknowledged_at`, `bugbot_down`, `copilot_down`, `agents_session_id`,
+`persistent_agents`.
 
 ## Persistent per-step agents
 
@@ -190,19 +189,17 @@ the `persistent_agents` map
   resolver-supplied sonnet-equivalent on third-party hosts. The `name`
   makes the agent a persistent teammate
   that idles awaiting messages. Record `{agent_id, created_tick,
-  last_used_tick}` under the step key. Keep the spawn prompt fix-shaped,
-  never audit-shaped: the `pr_converge_bugteam_enforcer` hook blocks
-  audit-shaped clean-coder spawns during the BUGTEAM phase.
+  last_used_tick}` under the step key. Keep the spawn prompt focused on the
+  current fix. Step 6 sends audit judgment through the
+  [review guide](../reviews/SKILL.md#review-workflow).
 - **Stale or dead id:** on a `SendMessage` failure, or no acknowledgment
   within one bounded wait, drop the map entry, spawn a fresh named agent,
   record it, and continue the tick. Never abort a tick on a stale id;
   never retry the same dead id.
 - **Fresh every round (never persisted):** the Step 5 host-aware
   `invoke_code_review.py` / `/code-review ultra --fix` pass and the Step 6
-  bugteam audit (unbiased eyes each round; the enforcer needs the formal
-  Skill call), and every `code-verifier` — a named code-verifier never fires
-  `SubagentStop`, so no verdict mints (see the named-`code-verifier` entry
-  in the Gotchas list below).
+  bugteam audit. Apply the [review guide](../reviews/SKILL.md#review-workflow)
+  to each review pass.
 - **Shutdown:** at loop end (convergence or a stop condition), send each
   persistent agent a shutdown request and clear `persistent_agents` before
   the `pr-loop-lifecycle` Close.
@@ -263,7 +260,6 @@ post a fresh PR in a fresh branch based on origin main to the user.
   [Step 1.5](reference/per-tick.md). Skipping this reviews and edits the
   wrong repo. The route is routine and automatic — never a material fork
   to pause on.
-- **A named/teammate `code-verifier` never mints a verdict** — In a background-job session, an `Agent`-tool `code-verifier` spawned with a `name` (or otherwise as a persistent teammate) goes idle/"available" awaiting messages rather than terminating, so its `SubagentStop` never fires. `verifier_verdict_minter.py` mints the verdict only on `SubagentStop`, so no verdict file is written and `verified_commit_gate` blocks the `git commit`/`git push` with "no passing verification verdict" even though the verifier emitted `all_pass`. A `shutdown_request` is ignored and `TaskStop` cannot resolve the teammate's id. Spawn the code-verifier as a one-shot agent with NO `name` (a plain async `Agent` call) so it runs to completion and fires `SubagentStop`, minting the verdict bound to the live surface. Keep the work tree frozen between verification and the commit so the minted surface hash still matches.
 
 ## Progress checklist
 
@@ -333,7 +329,7 @@ round as converged. This rule holds every tick, every loop, every PR.
       - [ ] **clean on `current_head`** → zero unresolved threads (else fix + resolve first)
             → `bugbot_clean_at = current_head` → Step 7
       - [ ] **no review / commit_id mismatch** → `reviewer-gates` Bugbot flow (Gate 3):
-            silent pass → stamp + Step 7; queued/triggered → ScheduleWakeup 360s or portable in-session poll → Step 4;
+            silent pass → Step 7; queued/triggered → ScheduleWakeup 360s or portable in-session poll → Step 4;
             down → `bugbot_down = true` → Step 7
 
 - [ ] **Step 5: CODE-REVIEW — static sweep, review, fix, advance**
@@ -373,9 +369,10 @@ round as converged. This rule holds every tick, every loop, every PR.
       See: [`reference/per-tick.md` § Step 2 BUGTEAM](reference/per-tick.md);
       [`../bugteam/SKILL.md`](../bugteam/SKILL.md).
       Pre-condition: `code_review_clean_at == current_head`.
-      Mandatory: `Skill({skill: "bugteam", args: "<PR URL>"})` this tick
-      (enforcer-blocked otherwise; `qbug` is not a substitute). Scope: FULL
-      `origin/main...HEAD` diff. Re-resolve HEAD after bugteam.
+      Mandatory: `Skill({skill: "bugteam", args: "<PR URL>"})` this tick.
+      Use the [review guide](../reviews/SKILL.md#review-workflow) for its
+      judgment. `qbug` is not a substitute. Scope: FULL `origin/main...HEAD`
+      diff. Re-resolve HEAD after bugteam.
 
       - [ ] **bugteam pushed** → verify threads replied + resolved → reset markers
             → `phase = CODE_REVIEW` → ScheduleWakeup 360s or portable in-session poll → Step 5
@@ -392,7 +389,7 @@ round as converged. This rule holds every tick, every loop, every PR.
       before these gates run. Count unresolved threads before each gate.
       Every gate records evidence; gate (f) cites evidence from (a)–(e).
 
-      - [ ] **(a) Copilot findings** — fetch Copilot on `current_head`; dirty → fix + return to Step 5; clean → stamp `copilot_clean_at`; absent → continue; when `copilot_down`, skip
+      - [ ] **(a) Copilot findings** — fetch Copilot on `current_head`; dirty → fix + return to Step 5; clean → record `copilot_clean_at`; absent → continue; when `copilot_down`, skip
       - [ ] **(b) Claude reviewer** — fetch Claude on `current_head`; dirty → fix + return to Step 5; clean or absent → continue
       - [ ] **(c) Mergeability** — `mergeable_state == "clean"` and `mergeable == true`; dirty → rebase + return to Step 1; blocked/behind/unknown/unstable → hard blocker
       - [ ] **(d) Post-convergence Copilot request** — request Copilot when not pending and not `copilot_down`; enter `COPILOT_WAIT` (Step 7a); when `copilot_down`, skip to (e)