Agent-Skills.md

Agent Skills: Evaluate Plugin

Audit a local Claude Code bundle that has .claude-plugin/plugin.json: validate the manifest, lint hooks, check mcp/lsp servers, monitors, and agents, then rank fixes. Use when the user says "audit this", "validate the manifest", "lint the hooks", "why did it score that way", or "benchmark this".

UncategorizedID: ANcpLua/ancplua-claude-plugins/evaluate-plugin

Repository

ANcpLua/ancplua-claude-plugins
ANcpLuaLicense: MIT
51

Install this agent skill to your local

pnpm dlx add-skill https://github.com/ANcpLua/ancplua-claude-plugins/tree/HEAD/plugins/cc-plugin-eval/skills/evaluate-plugin

Skill Files

Browse the full folder contents for evaluate-plugin.

Download Skill

Loading file tree…

plugins/cc-plugin-eval/skills/evaluate-plugin/SKILL.md

Skill Metadata

Name
evaluate-plugin
Description
'Audit a local Claude Code bundle that has .claude-plugin/plugin.json: validate the manifest, lint hooks, check mcp/lsp servers, monitors, and agents, then rank fixes. Use when the user says "audit this", "validate the manifest", "lint the hooks", "why did it score that way", or "benchmark this".'

Evaluate Plugin

Use this skill when the target is a plugin root with .claude-plugin/plugin.json. If the target is a single SKILL.md directory, hand off to ../evaluate-skill/SKILL.md instead.

The runnable commands below invoke the shipped script through the plugin root, which works in any checkout: node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" <command> <args>. Inline mentions of cc-plugin-eval <command> are shorthand for that same invocation; the bare cc-plugin-eval alias only exists if you ran npm link in the plugin directory.

When To Use

  • The path passed by the user contains .claude-plugin/plugin.json.
  • The user's question is about the bundle as a whole, not a single skill.
  • The user says "audit", "validate", "lint", "inspect", or "benchmark this plugin".

If the path is a SKILL.md or a directory holding one, route to ../evaluate-skill/SKILL.md. If the path contains neither, treat it as a generic directory and run cc-plugin-eval analyze <path> to surface the file-level signals.

Workflow

  1. Treat "Evaluate this plugin." as the default entrypoint.
  2. If the request comes in as natural chat language, run cc-plugin-eval start <plugin-root> --request "<user request>" --format markdown first so the user sees the routed local path.
  3. Run cc-plugin-eval analyze <plugin-root> --format markdown.
  4. Read Fix First before drilling into manifest findings, hook findings, MCP/LSP findings, monitor/agent findings, marketplace findings, then nested skill findings.
  5. If the plugin contains multiple skills, summarize the strongest and weakest ones explicitly.
  6. If the user wants a CI-friendly pass/fail check, run cc-plugin-eval validate <plugin-root> --strict. Exit code 2 means at least one warn or fail finding fired.
  7. If the user wants component-only output (only hooks, only mcp, etc.), run cc-plugin-eval inspect <plugin-root> --component <name>.
  8. If the user wants measured usage, switch to "Help me benchmark this plugin." and use the starter benchmark flow.
  9. If the user wants trend data, compare two JSON outputs with cc-plugin-eval compare before.json after.json.

Plugin-Specific Priorities

  • .claude-plugin/plugin.json validity: missing name (CC101), invalid name (CC102), invalid SemVer (CC103), env-var leak in metadata (CC110), files in .claude-plugin/ other than plugin.json (CC120), path-shape violations (CC130), missing path on disk (CC131).
  • Hooks: invalid event name (CC302), case-wrong event (CC303), invalid hook type (CC304), missing referenced script (CC305), ${CLAUDE_PLUGIN_ROOT} not used and not a system bin (CC306), mcp_tool referencing an undeclared server (CC310).
  • MCP servers: missing command (CC402), missing referenced script (CC403), relative command without ${CLAUDE_PLUGIN_ROOT} (CC404), suspicious secret in env (CC409).
  • LSP servers: missing command (CC502), missing extensionToLanguage (CC503), extension without leading dot (CC504), restartOnCrash without maxRestarts (CC508).
  • Monitors: duplicate name (CC603), invalid when syntax (CC604), on-skill-invoke:<name> referencing missing skill (CC605), ${user_config.X} for missing X (CC608).
  • Agents: forbidden hooks/mcpServers/permissionMode in plugin-shipped agents (CC703), invalid effort (CC705), invalid isolation (CC707), tool/disallowedTools overlap (CC709).
  • Marketplace alignment: plugin missing from plugins[] (CC803), version drift (CC804), source mismatch (CC805).
  • userConfig: invalid identifier (CC910), missing required fields (CC911), suspicious secret name without sensitive: true (CC915), channel server not in mcpServers (CC916).
  • Path traversal across all components (CC900).
  • Skill aggregate quality: delegated to ../evaluate-skill/SKILL.md.

Chat Requests To Recognize

  • Evaluate this plugin.
  • Audit this plugin.
  • Why did this score that way?
  • What should I fix first?
  • Validate the manifest.
  • Lint this plugin.
  • Inspect the hooks.
  • Inspect the mcp servers.
  • Audit the components.
  • Help me benchmark this plugin.
  • What should I run next?

Commands

node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" start <plugin-root> --request "Evaluate this plugin." --format markdown
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" analyze <plugin-root> --format markdown
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" validate <plugin-root> --strict
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" inspect <plugin-root> --component all --format markdown
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" inspect <plugin-root> --component hooks --format markdown
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" inspect <plugin-root> --component mcp --format markdown
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" inspect <plugin-root> --component agents --format markdown
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" start <plugin-root> --request "What should I run next?" --format markdown
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" compare before.json after.json
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" report result.json --format html --output ./cc-plugin-eval-report.html
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" init-benchmark <plugin-root>
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" benchmark <plugin-root>

Reference

  • ../../references/chat-first-workflows.md
  • ../../references/component-validators.md