audit-procedures
Health check procedures D1–D14 for the Audit agent — structural validation, attention budget, version checks, workspace integrity, and static audit
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Health check procedures D1–D14 for the Audit agent — structural validation, attention budget, version checks, workspace integrity, and static audit
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Configure and manage Model Context Protocol servers for external tool access
Review a UI for accessibility — WCAG 2.1 AA compliance, semantic HTML, ARIA usage, keyboard navigation, focus management, colour contrast, and screen reader compatibility
Design or review a REST or GraphQL API — resource modeling, versioning strategy, error contract, OpenAPI/schema-first workflow, and security baseline
Generate a CHANGELOG.md entry from staged changes, a commit range, or a PR diff — following Keep a Changelog format with conventional commit classification
Set up and audit environment variable management — create .env.example, add startup validation, separate secrets from config, and document every variable
Generate or update onboarding documentation — README, CONTRIBUTING guide, dev environment setup script, and new-developer validation checklist
| name | audit-procedures |
| description | Health check procedures D1–D14 for the Audit agent — structural validation, attention budget, version checks, workspace integrity, and static audit |
| compatibility | >=2.0 |
Skill metadata: version "1.0"; license MIT; tags [audit, health, structural-validation, workspace, version]; compatibility ">=2.0"; recommended tools [codebase, runCommands, fetch].
Detailed health check definitions for the Audit agent. Load this skill before running health checks to get the full D1–D14 check specifications, thresholds, and flag levels.
Print a structured health report with sections D1–D14 showing findings, "OK", or "N/A" when a check does not apply to the detected repo shape. End with counts for CRITICAL/HIGH/WARN/INFO/OK and an overall health status.
Developer template repo only.
Count total lines and per-section lines using wc -l and grep -n "^## §".
| Scope | Limit |
|---|---|
| Entire file | ≤ 800 |
| §5 Operating Modes | ≤ 210 |
| §1–§4, §6–§9 (each) | ≤ 120 |
| §10 | No limit |
| §11, §12, §13, §14 (each) | ≤ 150 |
Flag: [CRITICAL] if any section exceeds limit. [WARN] if within 10 lines.
Developer template repo only.
Verify §1–§14 all present and in order.
Flag: [CRITICAL] if missing. [WARN] if out of order.
.github/copilot-instructions.md must have zero {{ tokens → [CRITICAL] if found.template/copilot-instructions.md must retain ≥ 3 {{ tokens → [HIGH] if fewer.For each .github/agents/*.agent.md: always check frontmatter present,
name: set, handoff targets resolve to existing agent names, and model: is
listed.
Developer template repo only: specialist delegation allow-lists match the repo policy.
Consumer repos: skip repo-policy allow-list matching and only report structural agent validity.
Flag: [CRITICAL] broken handoff. [HIGH] missing name/frontmatter or
required delegates in developer template repos. [WARN] missing model or
unexpected delegates in developer template repos.
If present: verify owned servers (mcp-git-server, mcp-fetch-server, mcp-duckduckgo-server, mcp-docs-server, mcp-sequential-thinking-server) use uvx transport not npx. Verify no upstream references to mcp-server-fetch, mcp-server-git, @modelcontextprotocol/server-sequential-thinking, or duckduckgo-mcp-server (all replaced by owned copies). Verify github server uses /readonly URL with X-MCP-Toolsets header.
Flag: [CRITICAL] npx usage for owned servers. [HIGH] upstream server names still present. [MED] github server using full write URL.
Use the repo shape detection in the agent body.
Developer template repo: skip (mark N/A). Consumer repo: check
.github/copilot-version.md exists, starts with a valid semver, includes
Applied: and Updated: dates, and carries section-fingerprints,
file-manifest, and setup-answers blocks.
Flag: [HIGH] if absent or malformed. [WARN] if fingerprint tracking is absent.
Check each file under .copilot/workspace/ exists and is non-empty.
Exempt transient files from the non-empty check: any file matching *.lock or living under .copilot/workspace/runtime/ or .copilot/workspace/.tmp/ is a transient heartbeat artifact and may legitimately be zero-byte during or between sessions.
Flag: [HIGH] if HEARTBEAT.md or IDENTITY.md missing. [WARN] for others (excluding transient files).
Present? References .github/copilot-instructions.md?
Flag: [WARN] if absent.
Check .vscode/settings.json for chat.pluginLocations — verify each enabled
path resolves. Accept a list only as a legacy fallback.
Check for naming conflicts between .github/agents/ and plugin-contributed agents.
Check for skill name collisions between .github/skills/ and plugin-contributed skills.
Flag: [WARN] for conflicts or non-existent paths. Skip silently if no plugin settings.
Check if installed via code --list-extensions. If installed, verify it appears in
.vscode/extensions.json recommendations.
Flag: [INFO] if not installed. [WARN] if installed but missing from recommendations. Skip if code CLI unavailable.
Skip in developer repo. Fetch raw.githubusercontent.com/asafelobotomy/copilot-instructions-template/main/VERSION.md, compare against .github/copilot-version.md.
Flag: [HIGH] behind by major version. [WARN] behind by minor/patch. [INFO] up to date. [WARN] if fetch fails.
Skip in developer repo. Parse <!-- section-fingerprints --> block from .github/copilot-version.md. For each §1–§9, compute fingerprint via sha256sum of section content and compare against stored value.
Flag: [INFO] per drifted section. [WARN] if ≥ 5 of 9 sections drifted. [WARN] if fingerprint block absent.
Skip in developer repo. Fetch raw.githubusercontent.com/asafelobotomy/copilot-instructions-template/main/.copilot/workspace/operations/workspace-index.json. Verify local project has all expected agents, skills, prompts, instructions, hook scripts, hook config, the setup workflow, and core .copilot/workspace/ files. Inspect installed starter-kit payloads under .github/starter-kits/*/ when present, and verify .vscode/settings.json, .vscode/extensions.json, and .vscode/mcp.json when the corresponding consumer surfaces exist.
Flag: [HIGH] missing agent, shell hook, workflow, core workspace file, or installed starter-kit payload. [WARN] missing skill, prompt, instruction, or PS1 hook. [INFO] for user-added extras. [WARN] if fetch fails.
If scripts/copilot_audit.py exists, run the profile that matches the detected
repo shape and map findings to the report (CRITICAL→CRITICAL, HIGH→HIGH,
WARN→WARN, INFO→INFO).
python3 scripts/copilot_audit.py --root . --output jsonpython3 scripts/copilot_audit.py --profile consumer --root . --output jsonCovers: A1–A4 (agents), C1 (consumer companion completeness), I1–I4 (instructions), V1 (version metadata), P1 (prompts), S1–S2 (skills), M1–M3 (MCP), H1–H2 (hooks), SH1–SH3 (shell), PS1 (PowerShell), K1–K2 (starter kits), and VS1 (VS Code settings).
Consumer static-audit subset covers: A1–A3 (agents), C1 (consumer companion completeness), I1, I3, and I4 (instructions), V1 (version metadata), P1 (prompts), S1–S2 (skills), M1–M3 (MCP), H1–H2 (hooks), SH1–SH3 (shell), PS1 (PowerShell), K1–K2 (starter kits), and VS1 (VS Code settings). It intentionally skips repo-only A4.
If absent: [INFO] — static audit skipped.