agentleFS
Sign inStart free
fritzprixfritzprix/libr-agent/.agents/skills/code-audit-expert/SKILL.md

code-audit-expert

Skill11 starsChanged 18 days agoMITView on GitHub

Critically audit refactorings and design decisions against the real codebase, verify claims vs code, score only relevant quality axes, and write evidence-based tech reports. Use after refactors, when reviewing agent-reported work, or when asking for architecture / ISP / DRY / token-cost / reliability feedback. For builtin MCP tool audits, prefer critique-builtin-tool or lean-builtin-tool-auditor instead.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    No fabricated metrics: Do not invent token savings, coverage %, speedups, or “100% tests” unless you measured them in this session.

    line 24
  2. Very specific:

    Cite evidence: Prefer path:line or short quoted snippets.

    line 26
  3. Distinctive:

    Use when scoring; omit items that do not apply.

    line 32
  4. Very specific:

    When changing a core type or service (example: BaseAIService / src/lib/ai-service/base-service.ts), inspect call sites for correct type …

    line 44

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Code Audit Expert

Audit completed work against the codebase. Prefer skepticism over praise. Every claim needs a file/line or a measured number.

Scope routing

  • Builtin MCP tools (schema, hints, Tool Design Manifesto): use critique-builtin-tool or lean-builtin-tool-auditor.
  • This skill: architecture fit, claim verification, DRY/ISP, token/cost impact, side effects after refactors or feature work (TS/React, Rust, shared libs).

Workflow

  1. Gather context — Changed files, PR notes, agent summary. Treat summaries as claims, not facts.
  2. Verify against code — Open the actual implementations. Build a claim→evidence table (see template §2). Mark invents, mismatches, and unstated gaps.
  3. Analyze only what applies — Interface fit (ISP), duplication/abstraction (DRY), cost (tokens/caching), reliability/side effects. Skip irrelevant axes.
  4. Report — Follow reporting-template.md. Write under .libragent/work/ (e.g. .libragent/work/code_audit_report.md). Do not invent paths.

Hard rules

  • No fabricated metrics. Do not invent token savings, coverage %, speedups, or “100% tests” unless you measured them in this session. Prefer qualitative statements or cite the command/output.
  • Score only relevant axes. Irrelevant rows = N/A with one-line reason. Never force a 5/5 for flavor.
  • Cite evidence. Prefer path:line or short quoted snippets. Star ratings without citations are invalid.
  • Surface risks. Side effects, regressions, and remaining debt go in the report even when the change is otherwise good.
  • Language. Skill instructions are English. Match the user’s language for the written report (Korean user → Korean report is fine).

Critique checklist (inline)

Use when scoring; omit items that do not apply.

  • Claims in the agent/PR summary match the code
  • Public contracts (schemas, APIs, tool responses) stay coherent for consumers
  • Shared logic is extracted once (not copy-pasted) without over-abstraction
  • Hot paths / LLM context avoid redundant payload echo
  • Error handling and limits (size caps, LCS bounds, etc.) are explicit
  • Tests cover the new contract fields and failure modes that matter
  • Feature-flag / build-variant behavior is consistent where dual builds exist

Consumer / dependency check

When changing a core type or service (example: BaseAIService / src/lib/ai-service/base-service.ts), inspect call sites for correct type narrowing and broken assumptions—not only the edited file.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free