# Audit

> System Audit — entry point

- Skill: `pantyuhov9-web/audit` (Agent Skill)
- Install (CLI): `npx skillmds add pantyuhov9-web/audit`
- Raw SKILL.md: https://api.skillmd.com/api/skills/pantyuhov9-web/audit/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Security
- Author: pantyuhov9-web (https://skillmd.com/u/pantyuhov9-web)
- Updated: 2026-08-19
- Page: https://skillmd.com/skills/pantyuhov9-web/audit

---


# System Audit — entry point

This skill is the invocation handle for the two-layer maintenance loop. The full architecture, mechanical/conceptual check lists, verbatim Layer 2 subagent prompt, output format, and constraints all live in the workflow doc — this file just routes you there.

## When this fires

Auto-triggers on "audit", "check the system", "find issues", "repo health" — or explicit `/audit`. Designed for "leftover credits" use: 1–2× per day, find one improvement, fix it, done. Compounds without ever feeling like a project.

## Begin

Read [workflows/system_audit.md](../../../workflows/system_audit.md) — authoritative spec. The "Orchestration sequence" section defines the 5 steps and contains the verbatim Layer 2 prompt to pass to the subagent. The "What each layer catches" tables enumerate the 10 mechanical and 10 conceptual categories.

Then execute Steps 1–5 from the workflow. Hard rule from the workflow: **only ONE fix per invocation.**

## See also

- [tools/system_audit.py](../../../tools/system_audit.py) — Layer 1 implementation
- [.system_audit_ignore](../../../.system_audit_ignore) — suppression rules