agent-audit

name	agent-audit
description	Audit code-review agents, skills, and hooks for structural compliance. Use this when adding or modifying any agent, skill, or hook file, or for a periodic health check of the toolkit. Trigger phrases: "audit the agents", "check compliance", "validate the skills", "are the agents correct", or any time agent/skill files change.
argument-hint	[file-path | --all] [--fix]
user-invocable	true
allowed-tools	Read, Edit, Grep, Glob, Bash(python3 scripts/validate_agent_contract.py *, python3 scripts/check_registry_sync.py *, python3 plugins/dev-team/scripts/verify_tier.py *, python3 -m pytest tests/agents/test_agent_knowledge_anchor.py, python3 "${CLAUDE_PLUGIN_ROOT}/scripts/claude_setup_review.py" *, python3 "${CLAUDE_PLUGIN_ROOT}/scripts/token_efficiency_review.py" *)

Agent Audit

Role: orchestrator. This skill performs mechanical compliance checks — pattern matching against known-good structure.

You have been invoked with the /agent-audit skill. Audit agents and skills for compliance with the eval system patterns documented in .claude/docs/eval-system.md.

Monorepo-relative by design (#1637). Steps 8 and 2e's scripts/validate_agent_contract.py and scripts/check_registry_sync.py calls are intentionally bare, not a dangling-path defect. check_registry_sync.py hardcodes plugins/dev-team and checks this marketplace repo's own registry — it exists only at this repo's root, so no ${CLAUDE_PLUGIN_ROOT} path could ever resolve it. validate_agent_contract.py's repo-root copy is a thin wrapper delegating to plugins/marketplace-dev/scripts/validate_agent_contract.py (which does ship and is portable) — but there is no single variable spanning two independently-installed plugins, so the wrapper is reachable only from this monorepo checkout, where both plugin directories are siblings on disk. Both calls only make sense run from a dev-team-monorepo-shaped checkout (this repo, or a structural fork), never an installed plugin cache. See tests/repo/test_agent_implemented_by_resolves.py's INTENTIONAL_BARE_INVOCATION set, which checks the "does not ship" premise mechanically. (#1652 landed a Bash(...) grant in this skill's allowed-tools frontmatter covering all three of these script steps — including verify_tier.py below — by their exact bare/repo-relative invocation form, so they execute today; see the allowed-tools block above.)

Orchestrator constraints

1. Check structure, not semantics. Verify required sections, fields, and patterns exist. Do not evaluate whether detection rules are good — that's agent-eval's job.
2. Deterministic checks only. Every check should be reproducible: does the field exist? Is the format correct? Does the section match the expected pattern?
3. When --fix is used, apply minimal structural fixes. Insert missing sections/fields using templates. Do not rewrite existing content.
4. Be concise. Output the report table and action items. No preambles, no per-file narration, no restating what was checked.

Steps

1. Parse arguments

Arguments: $ARGUMENTS

• No argument or --all: audit everything
• A specific file path (e.g., .claude/agents/js-fp-review.md): audit that file only
• --fix: after generating the report, automatically apply fixes for FAIL/WARN items

2. Audit agents

Scope: this audit covers agents from all plugins in the repository, not just plugins/dev-team/agents/. Currently audited directories:

• plugins/dev-team/agents/ — the primary dev-team plugin agents
• plugins/security-assessment/agents/ — the security-assessment plugin agents

The automated effort-band gate (tests/agents/test_agent_effort_frontmatter.py) discovers and checks agents from both directories; add new plugin agent directories to AGENTS_DIRS in that test file when a new plugin is introduced.

Read each file in .claude/agents/*.md whose body contains a structured JSON output schema — either inline (a line with "status": "pass|warn|fail|skip") or a citation of knowledge/review-agent-output-contract.md — these are review agents. Check:

1. Structured output format: Does the agent specify a JSON output schema, inline or by citing the shared contract?

   • Review agents MUST include status, issues, and summary fields (inline or via the cited contract)
   • FAIL if a review agent has no output format and no contract citation

2. Severity definitions: Does the agent define severity levels?

   • MUST define error, warning, and suggestion with clear criteria
   • FAIL if severity levels are missing

3. Detection rules: Does the agent list what it detects?

   • MUST have a section listing specific patterns/issues to flag
   • WARN if detection rules are vague or missing

4. Scope boundaries: Does the agent declare what it ignores?

   • Review agents SHOULD state what other agents handle
   • WARN if missing (helps avoid duplicate findings)

5. Self-describing: Does the agent depend on external config?

   • Agents MUST NOT reference config/, review-config.json, or external config files
   • Thresholds, file scope, and defaults MUST be declared inline in the agent definition
   • FAIL if an agent references external config

6. File scope: Does the agent declare which file types it applies to?

   • Language-specific agents (e.g., js-fp-review) MUST declare their file scope
   • Language-agnostic agents (e.g., structure-review) may omit this
   • WARN if a language-specific agent has no file scope declaration

7. Skip support: Does the agent define when to return status: "skip"?

   • All review agents MUST have a ## Skip section
   • MUST describe conditions when the agent is inapplicable
   • MUST show the skip JSON response format
   • WARN if skip section is missing

8. Model/effort contract: Does the agent's frontmatter conform to the native Claude Code sub-agent contract (ADR 0026)?

   • Run python3 scripts/validate_agent_contract.py <file> and fold every into FAIL and every into WARN in the report. This is the sole frontmatter-contract check — required-field presence (, ), pattern, enum membership on any field present (, , , , , , , ), unknown/misspelled-key warnings, and the plugin-ignored-field warning for //.

2b. Audit agent tool declarations (all agents)

Read every file in .claude/agents/*.md (team agents and review agents). Check:

1. Skills-Skill invariant: If the agent body contains a ## Skills heading, the tools: frontmatter MUST include Skill.
   • The ## Skills section documents which skills an agent invokes. Without Skill in tools:, the agent cannot load skill content at runtime (when tools: is specified as an allowlist, only listed tools are available).
   • FAIL if a ## Skills section is present but Skill is absent from tools:.
   • PASS if no ## Skills section is present (the invariant does not apply).
   • PASS if ## Skills is present and Skill is in tools:.

Include the result in the agent report table under a Skills-Tool column.

Fix (when --fix is passed): Append , Skill to the tools: frontmatter line. Report FIXED: <agent> — Added Skill to tools:.

2. Code-intelligence MCP invariant (review agents): Every read-only *-review agent MUST grant the five code-intelligence MCP tools in tools: — mcp__codegraph__codegraph_explore and mcp__plugin_repowise_repowise__{get_context,get_symbol,search_codebase,get_risk} — so a review on a repo with a CodeGraph/Repowise index uses verified skeletons and resolved call graphs instead of raw whole-file reads (the grant is inert when the server is absent; agents fall back to Read/Grep/Glob). This mirrors the Skills-Skill invariant: it self-extends to future *-review agents.

   • FAIL if any *-review agent's tools: line is missing one or more of the five names.
   • PASS if every *-review agent grants all five.
   • Also checks that code-review/SKILL.md still contains the code-intelligence phrases (the five tool names and .codegraph/) as a proxy for the detection/preference guidance — it asserts the phrases are present, not the instruction wording itself.

3. Code-intelligence mapping invariant (non-review team agents): Every non-review team agent in the #1108 mapping MUST grant its tier's tools in tools: — the narrow tier (software-engineer, mutation-kill, qa-engineer, data-flow-tracer) grants codegraph + the four Repowise tools, with data-flow-tracer also carrying a scoped Bash(graphify *); the rationale tier (adr-author, architect, security-engineer, platform-engineer, codebase-recon) additionally grants mcp__plugin_repowise_repowise__get_why. Coverage is the union of the mapping's config keys and a structural sweep (team agents by ## Behavioral Guidelines or Enforcement: script, minus *-review, minus a documented exclusion list), so a new team agent that is neither mapped nor excluded fails rather than silently escaping the mapping. This is the non-review counterpart to invariant 2: review agents keep their five-tool set there; this check never evaluates agents.

Mechanics for invariants 2–4: all three delegate to plugins/dev-team/scripts/lib/mcp_tool_grants.py's shared run_grants_check (single read+parse pass per agent, both for detection and --fix) and, under --json, report the same envelope (check/evaluated/offenders/unclassified/fixed/unfixable/ok/notes) so results are comparable across the three checks.

Invariant	Report column	Delegated script	Config source of truth	Unclassified handling
2. Code-intelligence (review agents)	Code-Intel	plugins/dev-team/scripts/check_review_agent_mcp_tools.py	MCP_TOOL_NAMES	n/a — glob-discovered, no roster
3. Code-intelligence mapping (non-review)	Mapping	plugins/dev-team/scripts/check_agent_tool_mapping.py	TIER_CONFIG / EXCLUSIONS	reported, not auto-fixed
4. Code-intelligence (security-assessment)	SA-MCP	plugins/dev-team/scripts/check_security_assessment_mcp_tools.py	CODE_READING_AGENTS / NON_CODE_READING_AGENTS	reported, not auto-fixed

Include each invariant's result in the agent report table under its column above. Invariant 4 is additionally wired into scripts/ci-local.sh as chk_sa_mcp_tools, so drift fails CI, not just this manual audit pass.

All three delegated scripts above are repo-root-relative, not ${CLAUDE_PLUGIN_ROOT}-qualified — same working-tree-vs-cache reasoning as verify_tier.py above (§2b): each mutates agent files resolved relative to its own __file__, so a ${CLAUDE_PLUGIN_ROOT}-qualified invocation would silently edit the installed plugin cache's copy instead of the operator's own checkout.

Fix (when --fix is passed): run the delegated script above with --fix — it appends its missing tool names to tools: (merge, never replacing existing grants such as Read, Grep, Glob/Skill). Unclassified agents, where applicable, are reported, not auto-fixed — classify each into the config or exclusion list named above. Report FIXED: <agent> — added <tool names>.

2c. Audit team agent personas

A file is a team agent when its body contains a ## Behavioral Guidelines section. Exemption: an agent that declares Enforcement: script in its body is a script-enforced prose spec, not a persona-driven team agent — skip the persona checks below for it and instead verify it carries a > **Implemented by:** <script> pointer immediately after the H1. For each remaining team agent, check:

1. Persona paragraph: Is there a You are… sentence immediately after the H1 heading and before the first ## section?

   • The line must begin with You are (case-sensitive).
   • FAIL if the first non-blank line after the H1 is a ## heading instead of a persona paragraph.
   • PASS if a You are … paragraph is present between the H1 and the first ##.

2. Non-generic Output discipline: Does the ## Output discipline section exist and contain role-specific content?

   • FAIL if the ## Output discipline section is absent entirely.
   • WARN if the first bullet still contains the old generic phrase "plans, designs, ADRs, reports" — this indicates the shared boilerplate was never personalised.
   • PASS if the section is present and does not match the generic boilerplate.

Include both results in the agent report table under Persona and Output-Disc columns.

Fix (when --fix is passed):

• Missing persona paragraph: insert a placeholder You are a <role>. <Add identity, worldview, communication style.> after the H1. Report FIXED: <agent> — Added persona placeholder (requires manual completion).
• Missing ## Output discipline: insert the section with a placeholder bullet. Report FIXED: <agent> — Added Output discipline placeholder (requires manual completion).
• Generic boilerplate detected: emit WARN: <agent> — Output discipline still contains generic boilerplate; manual update required (no auto-fix — content must be role-specific).

2d. Citation drift lint (preventive)

Reviewer agents sometimes inline normative rules — numeric thresholds like "under 50 lines" or "80% coverage" — independently of the canonical skill or knowledge file. When that source changes, the agent silently keeps enforcing the stale value. The citation lint makes the dependency explicit: an agent declares its sources in a body-level Cites: list (not frontmatter — cites is not part of the official sub-agent contract), and every numeric threshold the agent states on an RFC-2119 line (MUST/SHOULD/SHALL/REQUIRED/NEVER/ALWAYS) must also appear in a cited source.

Perform the check by reading (the same mechanical, deterministic style as the other audits — no judgment):

1. Read each agent's body and look for a Cites: list. Each entry names a skill (skills/<name>/SKILL.md) or knowledge file (knowledge/<name>.md).
2. In the agent body, find every line carrying an RFC-2119 keyword (MUST/MUST NOT/SHOULD/SHALL/REQUIRED/NEVER/ALWAYS). Ignore lines inside code fences (```/~~~) and blockquotes (>). On each such line, collect the numeric thresholds (50, 80%, 40.5); ignore issue refs like #99.
3. Read each cited source and check the threshold appears in it.

Classify:

• Cites: present and every threshold backed → PASS in the Citation column.
• Cites: present but a threshold absent from every cited source → WARN (possible drift): report the token + line number.
• no Cites: but the agent states thresholds → WARN (advisory): recommend adding a Cites: list.
• no Cites: and no thresholds → PASS (nothing to verify).
• Cites: an unknown source (no matching skill/knowledge file) → WARN.

Phase 1 is non-blocking — these are warnings, never failures. Do not red-line the audit on a citation warning; surface it as an action item so drift is visible while Cites: adoption grows. CI runs the deterministic counterpart, scripts/citation_lint.py (also advisory, exit 0), on every PR.

2e. Registry completeness (preventive)

The catalog tables (knowledge/agent-registry.md for agents and agent-loaded skills; the plugin CLAUDE.md slash-command table for user-invocable skills) are hand-maintained. When an agent or skill is added or removed without updating the matching table, the catalog drifts and the orchestrator routes against a roster that no longer matches the filesystem.

Run the deterministic sensor:

python3 scripts/check_registry_sync.py

It asserts a bijection between plugins/dev-team/agents/*.md + plugins/dev-team/skills/*/SKILL.md and the registry rows, reporting MISSING (file with no row) and ORPHAN (row with no file). Unlike the citation lint this is a hard gate — exit 1 on any discrepancy. Fix it by adding or removing the catalog row by hand (the marketplace-dev plugin's /agent-create / /agent-remove maintain these tables automatically). Effort bands are deliberately not checked — they live only in frontmatter. The pytest suite tests/repo/test_registry_sync.py runs this on every PR.

2e-bis. Knowledge-reference integrity (preventive)

Agents read shared guidance from knowledge/*.md. Two ways this silently breaks (issue #1103): the file is renamed/removed but a reference lingers, or the reference uses a bare knowledge/X.md path that resolves against the target repo's cwd at runtime — not the plugin dir — so the agent reads nothing and degrades to hardcoded fallbacks.

Run the deterministic sensor:

python3 -m pytest tests/agents/test_agent_knowledge_anchor.py

It hard-gates three invariants on every knowledge/X.md reference in an agent body: it cites a valid index.json anchor or carries Whole-file load:; the file is packaged on disk; and it is prefixed with ${CLAUDE_PLUGIN_ROOT}/ (the only runtime-resolvable form — a prose "lives in plugins/dev-team/knowledge/..." pointer is tolerated). Runs on every PR.

2f. CLAUDE.md and token-efficiency structural checks

Two deterministic Python scripts validate concerns that the LLM-based review agents (claude-setup-review and token-efficiency-review) handle as deeper semantic checks. Run these scripts for a fast, CI-safe structural pass:

CLAUDE.md / setup review (frontmatter schema, field completeness, effort bands, duplicate rules):

python3 "${CLAUDE_PLUGIN_ROOT}/scripts/claude_setup_review.py" --plugin-root <plugin-root-path>

Token-efficiency review (file line counts, CLAUDE.md size, LLM anti-patterns):

python3 "${CLAUDE_PLUGIN_ROOT}/scripts/token_efficiency_review.py" --files <path>...

Both scripts exit 0 and emit JSON findings to stdout. Surface any error or warning severity findings as FAIL/WARN rows in the audit report table. The scripts are the authoritative structural gate — do not dispatch the claude-setup-review or token-efficiency-review agents from this skill; those agents run under /code-review for semantic depth.

3. Audit skills

Read each file in .claude/skills/*.md and .claude/skills/*/SKILL.md and check:

1. Role declaration: Does the skill declare its role?

   • User-invocable skills (user-invocable: true — slash commands and workflow initiators) MUST have an explicit Role: line in the SKILL.md body, immediately after the H1 (matching /plan, /build, /code-review, /pr, /ship): Role: orchestrator, Role: worker, or Role: implementation. A frontmatter-only role: field (lowercase key) does NOT satisfy this — it's easy to add without stating the orchestration-discipline contract the body line carries. WARN if a user-invocable skill has no body Role: line.
   • Agent-loaded, non-user-invocable knowledge skills (no user-invocable: true) are exempt from the body-line requirement — a frontmatter role: field, or no role at all for pure reference material, is acceptable. WARN only if such a skill has no role declared anywhere (frontmatter or body).
   • Orchestrators route work and aggregate results — they must not review or modify code
   • Workers perform semantic analysis using agent definitions
   • Implementation skills modify code following correction prompts

2. Constraints section: Does the skill declare its boundaries?

   • All skills SHOULD have a constraints section matching their role
   • Orchestrators: must not review code, must delegate, must minimize context
   • Workers: must follow agent definition, must return structured JSON
   • Implementation: must apply minimal fixes, must validate after changes
   • WARN if constraints are missing

3. Structured steps: Does the skill have numbered steps?

   • All skills MUST have a clear sequence of steps
   • FAIL if steps are missing or unstructured

4. Argument parsing: Does the skill document its arguments?

   • Skills MUST document required and optional arguments
   • WARN if argument section is missing

5. Output format: Does the skill describe its output?

   • Skills that produce reports MUST define their output format
   • WARN if output format is missing

4. Audit hooks

Read each file in .claude/hooks/*.sh and check (the *.sh glob already excludes .claude/hooks/freeze-state.json and .claude/hooks/careful-state.json — plugin-written runtime state, not operator-authored hook code; see docs/python-hook-contract.md § Environment variables, issue #1904 item 10):

1. Advisory behavior: Does the hook exit 0?

   • Hooks MUST be advisory only (exit 0), never blocking
   • FAIL if a hook exits non-zero on warnings

2. Input handling: Does the hook read stdin and extract file path?

   • Hooks MUST handle the PostToolUse input format
   • WARN if input parsing looks incorrect

3. Scope filtering: Does the hook filter by file type?

   • Hooks SHOULD only run on relevant file types
   • WARN if no file type filter is present

5. Generate report

# Agent Audit Report

## Agents
| Agent | Output Format | Severity | Detection | Scope | Self-Describing | File Scope | Skip | Model Tier | Context Needs | Skills-Tool | No-Colon Desc | Status |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| test-review | PASS | PASS | PASS | PASS | PASS | N/A | PASS | PASS | PASS | PASS | PASS | OK |
| js-fp-review | PASS | PASS | PASS | PASS | PASS | PASS | PASS | PASS | PASS | N/A | PASS | OK |
| ... | | | | | | | | | | |

## Skills
| Skill | Role | Constraints | Steps | Arguments | Output | Validation | No-Colon Desc | Status |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| code-review | PASS | PASS | PASS | PASS | PASS | N/A | PASS | OK |
| apply-fixes | PASS | PASS | PASS | PASS | PASS | PASS | PASS | OK |
| ... | | | | | | | |

## Hooks
| Hook | Advisory | Input | Scope Filter | Status |
| --- | --- | --- | --- | --- |
| js-fp-review.sh | PASS | PASS | PASS | OK |
| token-efficiency-review.sh | PASS | PASS | PASS | OK |
| ... | | | | |

## Citation drift (Phase 1 — advisory)
| Agent | cites | Drift / Advisory | Status |
| --- | --- | --- | --- |
| structure-review | yes | — | PASS |
| naming-review | no | states 1 threshold, no Cites: | WARN |
| ... | | | |

## Summary
- Agents: N OK, N WARN, N FAIL
- Skills: N OK, N WARN, N FAIL
- Hooks: N OK, N WARN, N FAIL
- Citation drift: N PASS, N WARN (advisory, non-blocking)
- Action items: [list of things to fix]

6. Apply fixes (when --fix is passed)

If --fix was NOT passed, list action items and stop.

If --fix WAS passed, automatically apply fixes for each FAIL/WARN item:

Agent fixes:

• Missing output format → insert after the # <Agent Name> heading:

  Output JSON: per `${CLAUDE_PLUGIN_ROOT}/knowledge/review-agent-output-contract.md` (Whole-file load: short, canonical schema).
  

  Prefer this citation form over inlining the schema — see knowledge/review-agent-output-contract.md for the canonical shape and its documented per-agent extensions/status exceptions.

• Missing severity definitions → insert after the output format:

  Severity: error=<agent-specific>, warning=<agent-specific>, suggestion=<agent-specific>
  

• Missing skip support → insert a ## Skip section before ## Detect:

  ## Skip
  
  Return `{"status": "skip", "issues": [], "summary": "<reason>"}` when:
  - <agent-specific inapplicability conditions>
  

• Missing scope boundaries → append ## Ignore section at the end

• Colon in description → rewrite the description value to remove colons (use "–" or "and" instead)

Skill fixes:

• Missing numbered steps → restructure existing content under ## Steps with ### 1., ### 2., etc.
• Missing argument section → insert ## Parse Arguments section after the skill heading
• Colon in description → rewrite the description value to remove colons (use "–" or "and" instead)

After each fix:

1. Read the file to confirm the fix was applied
2. Re-run the specific check to verify it now passes
3. Report: FIXED: <agent/skill> — <what was fixed>

7. Fix summary

If --fix was used, append a fix summary after the audit report:

## Fixes Applied
- FIXED: <name> — Added output format
- FIXED: <name> — Added skip section
- SKIPPED: <name> — <reason fix could not be auto-applied>

Re-run /agent-audit to verify all fixes.