vc-audit-vc

Validates agent harness consistency across Claude, Codex, README, and protocol files.

2|Updated Sep 15, 2025
One-click install
npx skills add https://github.com/marsley01/Marsley-Portfolio-Web --skill vc-audit-vc-marsley01
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: vc-audit-vc
Source: https://github.com/marsley01/Marsley-Portfolio-Web/tree/main/.agents/skills/vc-audit-vc
Command: npx skills add https://github.com/marsley01/Marsley-Portfolio-Web --skill vc-audit-vc-marsley01

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve? Multi-agent development setups drift over time: Claude and Codex agent definitions fall out of sync, skill registries miss entries, README documentation diverges from disk, and protocol files reference paths that no longer exist. This Skill runs a battery of validators to detect and report that drift before it breaks agent behavior. ## Core Features & Use Cases - Agent Parity Validation: Compares Claude markdown agents against Codex TOML mirrors for missing files, description mismatches, and mode inconsistencies. - Skill Registry & README Sync: Verifies every skill on disk has valid frontmatter, appears in the README catalog, and that the .agents/skills symlink resolves to .claude/skills. - Protocol & Wiring Checks: Confirms protocol files are listed in the router document, agent bodies reference only registered skills, and agent frontmatter includes required fields like effort and skills. - Use Case: After renaming a skill or adding a new agent, run this audit to get a JSON report of every broken reference, missing mirror file, or stale README entry, then patch and re-run until clean. ## Quick Start Run the vc-audit-vc skill to audit the agent harness and report any parity, registry, README, or protocol wiring inconsistencies.

Frequently Asked Questions about vc-audit-vc

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I check that Claude and Codex agent definitions stay in sync?

Run the validate-agent-parity.mjs script, which compares .claude/agents markdown files against .codex/agents TOML mirrors. It reports missing counterparts, description mismatches, and mode tag differences as JSON failures or warnings.

How to validate that all skills are listed in the README?

Run validate-guide-sync.mjs from the repository root. It extracts agent and skill names from README tables and inline backticks, then cross-checks them against SKILL.md files on disk, failing on any disk skill missing from the README catalog.

What does the skill invocation wiring validator check?

It scans agent bodies for backtick-wrapped vc-* tokens and verifies each one matches a registered skill in generated-skills-catalog.json. Known agent names are exempted via an allowlist, and unregistered references cause an exit code of 1.

Does this audit work outside a git repository?

Yes, most validators fall back to the current working directory when git rev-parse fails. However, scripts that import shared utilities from sibling skill directories still require the expected .claude/skills folder structure to resolve imports.

Why does the seeds validator warn when process/_seeds is missing?

The process/_seeds directory is an optional legacy scaffold surface in the live repository. Its absence is a warning-only result unless you are explicitly auditing export-kit scaffolding, in which case use --strict to promote warnings to failures.

When should I use audit-context instead of this audit?

Use the audit-context skill for context routing, grouping, and discoverability audits. This skill covers harness wiring only: agent parity, skill registry, README sync, protocol references, seeds, and kit portability.