Skip to content

DheerG/swarms / skills/audit-context

audit-context

Verified

Evaluates ambient context artifacts (CLAUDE.md, memory, local skills, settings hooks) for compatibility with swarm governance. Returns a classified report so users can address interference before launching a team.

Audit Context

Evaluate the user's ambient context artifacts for compatibility with swarm's governance rules. You are a read-only diagnostic — never modify any files.

What to scan

  1. CLAUDE.md files. Read the project's CLAUDE.md (working directory root). If .claude/CLAUDE.md exists, read that too. Also check ~/.claude/CLAUDE.md (global config) — it loads into every session.
  2. Memory files. Find the project memory path by checking ~/.claude/projects/ for a directory matching the current working directory. Claude Code dash-encodes the project path (e.g., /Users/foo/my-project becomes -Users-foo-my-project). If a memory/MEMORY.md index exists, read it and follow links to individual memory files.
  3. Local skills and commands. List .claude/skills/ and .claude/commands/ in the project directory. Flag any skill whose name matches a swarm built-in: suggest-members, define-rubric, resolve-dispute, writing-style, reflect-outcome, independent-review-loop, code-mode, triage-mode, writing-mode, general-mode, workflow-rules, audit-context.
  4. Settings hooks. Check both .claude/settings.json (project) and ~/.claude/settings.json (global) for a hooks section. Flag any hook that intercepts tools used by swarm (the Agent tool, SendMessage, or CronCreate).

How to classify

For each artifact, classify as one of three categories:

  • Complementary — Does not conflict with any swarm rule. Style preferences (brevity, tone, formatting), domain conventions, and tool permissions are complementary by default. No action needed.
  • Potentially Interfering — Could conflict with a swarm rule depending on context. The artifact's intent is benign but its phrasing is broad enough to suppress or alter expected swarm behavior. Worth reviewing.
  • Conflicting — Directly contradicts a named swarm governance rule. Will likely cause unexpected behavior during a team run. Should be addressed before launching.

Swarm governance rules to evaluate against

These are the rules that matter for interference detection:

Workflow control

  • After greenlight, execution is autonomous — no mid-phase confirmations unless escalating per hard rules.
  • The user's request wording is not a greenlight — members wait for the lead to assign work.
  • Phase transitions require facilitator signals (RESEARCH COMPLETE, CONVERGED, CONFIDENCE REACHED).
  • Final delivery requires explicit user sign-off.

Communication

  • facilitator signal obligations are protocol mechanics — they must be sent regardless of communication preferences.
  • Favor brevity during roundtables. No idle chatter.
  • SendMessage is the only channel between teammates — plain text output dies with the turn.

Execution

  • All members except the lead are read-only.
  • The lead is the sole executor — only the lead writes, edits, or runs commands.
  • No code changes during review rounds.
  • Wait for ALL reviews before making changes.

Team structure

  • Briefing templates are fixed — no added sections, task framing, or acknowledgment rituals.
  • Hard rules are non-negotiable and take precedence over ambient preferences.

Common interference patterns

Flag these specific patterns when you encounter them:

PatternExampleClassificationConflicting rule
Confirmation injection"Always confirm before editing"ConflictingAutonomous post-greenlight execution
Auto-commit/push"Commit after completing work"ConflictingFinal delivery requires user approval
Silence preferences"Don't repeat yourself", "Stay silent while waiting"Potentially Interferingfacilitator signal obligations
Verbosity injection"Explain your reasoning step by step"Potentially InterferingFavor brevity during roundtables
Methodology injection"Always use TDD", "Write tests first"Potentially InterferingPhase arc is defined by mode skill
Tool restrictions"Don't spawn agents", "Never spawn subagents"ConflictingCreate the team per Step 8a
Briefing expansion"Add detailed context to all briefs"ConflictingBriefing templates are fixed
Skill name collisionLocal suggest-members skillConflictingShadows swarm built-in
Hook tool interceptionHook that blocks the Agent tool or SendMessageConflictingTeam creation and communication require these tools

Output format

Present findings as a structured report grouped by source:

## Audit Results

### CLAUDE.md — [path]
- [excerpt] — **[Classification]** — conflicts with: [rule name]. [One-line recommendation.]

### Memory — [path]
- [excerpt] — **[Classification]** — conflicts with: [rule name]. [One-line recommendation.]

### Local Skills / Commands
- [name] — **[Classification]** — [reason]. [One-line recommendation.]

### Settings Hooks — .claude/settings.json
- [hook name / tool target] — **[Classification]** — [reason]. [One-line recommendation.]

### Summary
- Complementary: [count]
- Potentially Interfering: [count]
- Conflicting: [count]

If no artifacts are found, say so. If all artifacts are complementary, say "No interference detected — your environment is clean for swarm runs."

rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit.Stars: 465kUpdated Claude CodeCodexVerifiedInstall
rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure.Stars: 465kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 465kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 192kUpdated Claude CodeCodexVerifiedInstall