Vigiles - Setup Guide¶
Source: skills.sh (12.3K combined installs) GitHub: zernie/vigiles (15⭐, MIT) Category: Agent Quality & Harness Verification First Seen: June 6, 2026 (skills.sh) Quality Tier: 🟢 Production (Gen Agent Trust Hub Pass / Socket Pass / Snyk Pass)
"Like Lighthouse for your agent harness" - Vigiles verifies that your CLAUDE.md/AGENTS.md, skills, and hooks are real and tested, then hardens them. The core loop: strengthen scans spec files for guidance() rules and proposes enforce() replacements backed by actual linter rules (free wins auto-applied, costly changes presented as tradeoffs); verify-docs-findable checks agents can actually discover the docs; test-harness and debug-my-harness validate the harness itself; dogfood-cli closes the loop by testing your own tooling. A governance-first cluster in the same family as CorpusIQ's internal audit skills.
Installation¶
npx skills add zernie/vigiles --skill strengthen
npx skills add zernie/vigiles --skill verify-docs-findable
npx skills add zernie/vigiles --skill test-harness
npx skills add zernie/vigiles --skill deep-research
Core Skills¶
| Skill | Installs | Use For |
|---|---|---|
| strengthen / edit-spec / adopt-spec / migrate-to-spec | 1.3-1.5K | Spec lifecycle: create, edit, adopt, and harden rules |
| test-harness / debug-my-harness | 1.5K / 973 | Validate and debug the agent harness itself |
| linter-docs / add-a-linter / generate-rule / pr-to-lint-rule | 30-1.4K | Linter rule management from docs and PRs |
| verify-docs-findable / review-docs / audience-check | 250-534 | Documentation discoverability and audience fit |
| deep-research | 670 | Structured deep-research workflow |
| audit-feedback-loop / enforce-rules-format / code-quality | 298-344 | Quality gates and rule formatting |
| screenshot / generate-logo / landing-site | 205-317 | Visual assets for tooling and sites |
| dogfood-cli | 30 | Test your own CLI like a user |
Prerequisites¶
- None - prompt-skill suite; no API keys or services
- A spec/AGENTS.md file in the target repo unlocks the strengthen/adopt-spec workflows
CorpusIQ Use Cases¶
- Agent-harness hygiene - verify-docs-findable and test-harness mirror CorpusIQ's own system-auditor and stack-doctor sweeps, as a cross-check methodology
- Governance hardening - the guidance → enforce upgrade loop is a concrete pattern for CorpusIQ's governance baselines (turn prose rules into enforced checks)
- Docs findability - audience-check and review-docs pair with CorpusIQ's docs SEO/AEO/GEO passes
Integration with CorpusIQ¶
CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.
This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.
Limitations / Verification¶
- Small repo (15⭐, 12.3K installs) - early-stage publisher; skills are stable but the ecosystem around them is thin
- Vigiles targets Claude Code conventions (CLAUDE.md/AGENTS.md); adapt paths when applying to Hermes profiles
- Verify install:
npx skills list | grep -E "strengthen|test-harness"
Security¶
Gen Agent Trust Hub: Pass · Socket: Pass · Snyk: Pass