Agent Skills: cc-plugin-eval

cc-plugin-eval

Use this as the umbrella entrypoint for local Claude Code plugin and skill evaluation. cc-plugin-eval is deterministic, local-first, and zero-dependency.

The runnable command blocks 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.

Start Here

1. Resolve whether the target path is a skill (contains SKILL.md), a plugin (contains .claude-plugin/plugin.json), or another local folder.
2. If the user spoke naturally, prefer the chat-first router so they see the routed local command:

node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" start <path> --request "<user request>" --format markdown

3. Map natural chat requests to workflows:
   • "Evaluate this plugin." -> cc-plugin-eval analyze <path> --format markdown
   • "Evaluate this skill." -> cc-plugin-eval analyze <path> --format markdown (or evaluate-skill alias)
   • "Why did this score that way?" -> cc-plugin-eval analyze <path> --format markdown
   • "What should I fix first?" -> cc-plugin-eval analyze <path> --format markdown
   • "Explain the token budget for this plugin." -> cc-plugin-eval explain-budget <path> --format markdown
   • "Validate the manifest." / "Lint this plugin." -> cc-plugin-eval validate <path> --format markdown
   • "Inspect the hooks." / "Inspect the mcp servers." -> cc-plugin-eval inspect <path> --component <name> --format markdown
   • "Measure the real token usage of this skill." -> init-benchmark then benchmark, then analyze --observed-usage, then measurement-plan
   • "Help me benchmark this plugin." -> starter benchmark flow
   • "What should I run next?" -> cc-plugin-eval start <path> --request "What should I run next?" --format markdown
4. If the user wants a rewrite plan, route to ../improve-skill/SKILL.md. That skill produces a brief and then hands the brief to skill-creator for the actual rewrite pass.
5. If the user wants a custom rubric, route to ../metric-pack-designer/SKILL.md.
6. If the user names a plugin or skill instead of giving a path, resolve it locally before running commands. Prefer ~/.claude/plugins/cache/<plugin-id> for installed plugins, then any repo-local plugins/<name> or skills/<name> directory. If still ambiguous, ask one short clarifying question.
7. When the request sounds like "give me a full analysis" rather than just "evaluate", do the fuller path:
   • run analyze
   • initialize .cc-plugin-eval/benchmark.json
   • surface the setup questions that refine the starter scenarios
   • preview the first benchmark command the user can execute next

Chat Requests To Recognize

• Evaluate this plugin.
• Evaluate this skill.
• Give me an analysis of this plugin.
• Why did this score that way?
• What should I fix first?
• Explain the token budget for this plugin.
• Validate the manifest.
• Lint this plugin.
• Inspect the hooks.
• Inspect the mcp servers.
• Audit the components.
• Measure the real token usage of this skill.
• Help me benchmark this plugin.
• What should I run next?

Matching Commands

node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" start <path> --request "Evaluate this plugin." --format markdown
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" start <path> --request "Give me a full analysis of this plugin, including benchmark setup." --format markdown
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" analyze <path> --format markdown
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" validate <path> --format markdown
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" validate <path> --strict
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" inspect <path> --component all --format markdown
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" inspect <path> --component hooks --format markdown
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" explain-budget <path> --format markdown
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" measurement-plan <path> --format markdown
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" init-benchmark <path>
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" benchmark <path>
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" improve <path> --brief-out ./skill-brief.json
node "${CLAUDE_PLUGIN_ROOT}/scripts/cc-plugin-eval.js" compare before.json after.json --format markdown

Output Expectations

• Treat the JSON payload as the source of truth. Markdown and HTML are renderings.
• Lead with At a Glance, Why It Matters, Fix First, and Recommended Next Step.
• Keep the why content terse and easy to skim.
• Call out whether budget numbers are static estimates or measured benchmark results.
• Show the user the exact chat phrase they can reuse next, the cc-plugin-eval start command that routes it, and the first local workflow command behind it.
• When the user asks for a full analysis of a named plugin, do not stop at the report if benchmark setup is still missing.
• Hand off to ../evaluate-plugin/SKILL.md for plugin-specific reports and to ../evaluate-skill/SKILL.md for skill-specific reports.

When To Hand Off

• For a skill rewrite plan: ../improve-skill/SKILL.md. That skill writes a brief and then hands the brief to the skill-creator skill (Anthropic's skill-creator@claude-plugins-official). cc-plugin-eval focuses on structural and budget signals; skill-creator owns LLM-graded per-skill rewrites.
• For a custom rubric or domain-specific check: ../metric-pack-designer/SKILL.md.
• For plugin-specific evaluation: ../evaluate-plugin/SKILL.md.
• For skill-specific evaluation: ../evaluate-skill/SKILL.md.

References

• ../../references/chat-first-workflows.md
• ../../references/technical-design.md
• ../../references/evaluation-result-schema.md
• ../../references/component-validators.md