audit · liujh0729-jpg — Claude Code skill: what it does, how to install
audit is a Claude Code skill that audits codebase subsystems for security vulnerabilities, correctness bugs, and maintainability debt. Activates during full audit prompts or via /audit.
Apache-2.0 ★ 515 updated 2026-11-17
npx skills add liujh0729-jpg/AI-enterprise-OEM --skill audit -a claude-codeWhat audit does
This skill analyzes the ccch1mneyyy/dsh-Go codebase or selected subsystems by mapping runtime paths from entry points down to teardown handlers. It keeps files untouched on findings-only runs, executing code modifications solely when asked.
During inspection, the agent follows external inputs across session APIs, bridge routes, SSH or relay connections, plugin permissions, and file access boundaries. It matches dependencies against vulnerability advisories and verifies caller registries before flagging obsolete code.
Discovered issues are grouped by severity with triggers, exact file locations, impact assessments, and remediation steps. Unverified findings stay separate from confirmed defects, preventing false assumptions across uninspected project areas.
What it helps with
- trace untrusted inputs across bridge routes and terminal session APIs
- detect security flaws in host connections across local engine, SSH, and relay
- check dependencies and lockfiles against security advisory records
- document verified defects with reproducible triggers and concrete fixes
Example prompts for the agent
- /audit
- Audit the host connection lifecycle and relay routes for security risks
- Run a read-only audit on session APIs and plugin permission boundaries
How to install audit
1Skills CLI (Vercel)
npx skills add liujh0729-jpg/AI-enterprise-OEM --skill audit -a claude-code2Manually: put the files in place
~/.claude/skills/audit/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/audit/
1Skills CLI (Vercel)
npx skills add liujh0729-jpg/AI-enterprise-OEM --skill audit -a codex2Manually: put the files in place
.agents/skills/audit/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add liujh0729-jpg/AI-enterprise-OEM --skill audit -a cursor2Manually: put the files in place
.cursor/skills/audit/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install liujh0729-jpg/AI-enterprise-OEM mobot-app/.agents/skills/audit2Manually: put the files in place
.github/skills/audit/SKILL.md1Skills CLI (Vercel)
npx skills add liujh0729-jpg/AI-enterprise-OEM --skill audit -a gemini-cli2Platform CLI
gemini skills install https://github.com/liujh0729-jpg/AI-enterprise-OEM --path mobot-app/.agents/skills/auditManually: put the files in place
.devin/skills/audit/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about audit
How to install audit in Claude Code?
Run npx skills add liujh0729-jpg/AI-enterprise-OEM --skill audit -a claude-code or save the definition file to ~/.claude/skills/audit/SKILL.md.
What does the audit skill do?
It examines runtime execution paths, entry points, resource bounds, and external inputs, outputting a structured severity report with remediation steps.
When should audit be used instead of review or vuln-check?
Use audit for repository-wide or subsystem-level evaluations; switch to review for specific pull changes and vuln-check for dedicated vulnerability assessments.
Source text
SKILL.md · Apache-2.0
--- name: audit description: Audit the ccch1mneyyy/dsh-Go repository or a broad subsystem for security, correctness, and maintainability risks. Use for audit requests or /audit; use review for a specific change and vuln-check for a security-only assessment. --- Establish the audit scope, then trace the relevant runtime paths. For a findings-only request, leave code unchanged; carry out fixes when the user has requested them. 1. Identify entry points, state owners, external inputs, and teardown paths within the requested scope. 2. Follow inputs through validation and consumers. In this app, inspect applicable host connections (local engine, SSH, relay), session APIs, bridge routes, plugin capabilities, file access, permission flows, and long-session resource bounds. Prioritize paths whose failure can lose state, cross a trust boundary, or leave the terminal unusable. Check dependencies and lockfiles for known vulnerabilities against advisory data. 3. Confirm suspected defects against callers, existing guards, and focused regressions. Check exports, registries, and supported compatibility paths before declaring code dead. Explain a simplification in terms of current requirements and the behavior it preserves. 4. Report confirmed findings by severity with location, trigger, impact, and a concrete remedy. Keep unverified leads separate. For a cross-layer failure, a short input → boundary → effect trace may be enough to explain it. 5. State the areas examined, checks actually run, and material gaps. No findings means none found in that scope, not proof that the whole project is safe. An audit can finish without findings or a forced list of healthy areas.
© liujh0729-jpg · Apache-2.0 · liujh0729-jpg/AI-enterprise-OEM
Similar skills
All skills for Security →audit by other authors comes first.