Guard Skills: Guard Skills for AI Coding Agents
Second-pass quality gates that catch AI-generated failure modes in code, tests, and docs before they ship.
- Source repo
- amElnagdy/guard-skills
- Stars
- ★ 1.3k
- Last updated
- 3mo ago
- License
- MIT
- FA score
- 64/100 · Some gaps
At a glance
- How it runs
- Works with
- Platform-specificChatGPT · Codex · Claude Code
- Cost
- Free, no paid service needed
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A developer asks their coding agent to review a diff with
$clean-code-guardafter producing production code, to catch over-abstraction and error-swallowing before merge. - Not a fit if
- Teams wanting a full development workflow, not just review gates
- Teams expecting enforced automated checks; skills are guides, not checkers
- Environments without network or the Skills CLI, needed to install and update
- Source review
- 64/100 · Some gaps
What does this agent do, and when should you use it?
Guard Skills is a collection of focused guard skills for coding agents, acting as second-pass quality gates to catch systematic failure modes in AI-generated code, tests, and docs before delivery. It includes five skills: clean-code-guard (code quality gate for any language), test-guard (test code quality gate), docs-guard (documentation accuracy gate), wp-guard (WordPress-specific gate), and woo-guard (WooCommerce-specific gate). Each skill is structured with a lightweight SKILL.md entrypoint and progressive-disclosure references, with metadata supporting agents like Claude Code, Codex, Cursor, and OpenCode. Installation and updates are managed via the Skills CLI from Vercel Labs, supporting single-skill or global installation. The skills are best used as reactive review passes after an agent produces work, but can also guide writing when explicitly requested. The project is MIT-licensed, designed to be inspectable, containing only Markdown and YAML metadata with no executable scripts or credentials.
Guard Skills are distributed via the Skills CLI (npx skills add amElnagdy/guard-skills) and include five skills: clean-code-guard applies Clean Code, SOLID, DRY/KISS/YAGNI principles plus an AI-specific layer (catch-all error swallowing, hardcoded success returns, hallucinated APIs, premature abstraction, comment pollution, copy-from-similar bugs); test-guard enforces nine universal rules (e.g., mock only at system boundaries, parametrize instead of copy-pasting) with framework-specific progressive-disclosure references for pytest, PHPUnit/Pest, Jest/Vitest, Go tests, etc.; docs-guard treats documentation as a list of claims and verifies each against the codebase to catch hallucinated symbols and broken samples; wp-guard enforces WordPress practices (escaping, sanitization, nonces, capabilities, prepared queries, i18n, query/caching discipline); woo-guard enforces WooCommerce-specific rules (HPOS-safe order access, CRUD over direct meta, truthful feature-compatibility declarations, server-side checkout validation, money-handling discipline). Each skill is loaded via a SKILL.md entrypoint with deeper guidance loaded on demand.
- A developer asks their coding agent to review a diff with
$clean-code-guardafter producing production code, to catch over-abstraction and error-swallowing before merge. - After writing tests, a developer uses
$test-guardto check for mock abuse, duplicate tests, and assertions that don't catch anything. - Before shipping a README update, a developer runs
$docs-guardto verify all referenced symbols and examples are real and runnable. - A WordPress developer uses
$wp-guardto enforce escaping, nonce checks, and prepared queries in plugin changes. - A WooCommerce developer uses
$woo-guardto ensure HPOS compatibility and proper money handling in checkout customizations.
How do you install or deploy this agent?
- List available skills:
npx skills add amElnagdy/guard-skills --list. - Install the package:
npx skills add amElnagdy/guard-skills. - Install a single skill, e.g.,
npx skills add amElnagdy/guard-skills --skill clean-code-guard. - Install for a specific agent, e.g.,
npx skills add amElnagdy/guard-skills --skill test-guard --agent claude-code. - Install globally:
npx skills add amElnagdy/guard-skills --global. - Update skills:
npx skills updateor re-run the add command.
How do you use this agent?
Run a guard after your agent produces work: Use $clean-code-guard on the diff you just produced. Use $test-guard on the tests you just wrote. Use $docs-guard on the README update before shipping. Use $wp-guard on WordPress plugin changes. Use $woo-guard on WooCommerce checkout changes. You can also use a guard up front: 'Use $wp-guard while implementing this REST endpoint, then self-check before delivery.'
What are this agent's strengths and limitations?
- Specifically targets AI-generated code failure modes (error swallowing, hallucinated APIs) unlike generic linters.
- Uses progressive disclosure: lightweight SKILL.md loads cheaply, deeper guidance only when needed.
- Covers niche domains like WordPress and WooCommerce with platform-specific security and correctness rules.
- Inspectable: no scripts, credentials, or network calls, making it safe to audit.
- Requires Skills CLI and network access for installation, which may be a dependency.
- Skills are copied locally, so updates need manual re-installation; otherwise they may become stale.
- Limited to review gates, not a full development workflow, so it won't cover whole process.
- No guarantee that all best practices are covered; each skill is a guide, not a formal checker.
How does this agent compare with similar options?
Compared to WordPress/agent-skills which teach agents to build across the WordPress ecosystem, Guard Skills is narrower: it provides review gates to run after work is produced, focusing on catching common AI failure modes rather than building abilities.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Guard Skills: Guard Skills for AI Coding Agents This agent | 64 · Some gaps | CLIFree | ★ 1.3k | 3mo ago | — | ChatGPT · Codex · Claude Code |
| Agent Skills — Production-Grade Engineering Skills for AI Agents | 71 · Some gaps | CLIFree + model costs | ★ 108k | 5d ago | JavaScript | Codex · Claude Code · OpenAI API |
| ECC Engineering Harness | 66 · Some gaps | Agent plugin / skillFreemium | ★ 283k | 3d ago | JavaScript | Codex · Claude Code |
| Plannotator | 78 · Good | CLIFreemium | ★ 9.5k | today | TypeScript | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows: README explicitly states no executable scripts, no network calls, no MCP dependencies, no credentials; skill content is Markdown and lightweight metadata; installation requires user-initiated commands; updates require manual trigger; external URLs centralized in references/sources.md; MIT license. Deductions: no detailed permission inventory or data flow diagram; user confirmation mechanism not detailed; dependency security lacks specific audit info; rollback only mentions update commands; source attribution relies solely on repository metadata.
Evidence shows: README describes each skill's purpose and trigger, structure is consistent, dependencies are Skills CLI and external references, failure messages not detailed but skills designed as review feedback. Deductions: no specific examples of failure messages or error handling strategies; dependency availability unverified.
Evidence shows: clear target audience of coding agents, scenarios include code, tests, docs, WordPress, WooCommerce; capability boundaries clear; trigger methods explicit; supports multiple agents and environments. Deductions: no environment compatibility matrix or detailed configuration instructions.
Evidence shows: README provides install, update, usage instructions; skill naming clear; examples provided; MIT license; maintenance responsibility clear. Deductions: no version number or changelog; known limitations not explicitly listed.
Evidence shows: output is review feedback, directly usable for improvement; value in catching AI failure modes; low cost (no execution dependencies). Deductions: no actual effectiveness data or user feedback.
Evidence shows: README references research, external URLs in sources.md, facts and inferences separated. Deductions: no specific research citations or verification methods; cross-source corroboration insufficient.
- Static review; no skills executed; actual effectiveness unverified.
- Relies on Skills CLI; ensure its security and availability.
- External reference URLs unverified; may be stale or misleading.
- No versioning; updates may introduce breaking changes.
FAQ
Is there a cost to use these skills?
How do I update the skills after installing?
npx skills update or re-run the install command to re-fetch the latest version.