Use before creating or editing any SKILL.md files, and immediately after making skill changes to verify quality. Invoked when user asks about skill structure, descriptions, or best practices. Provides expert guidance on naming, descriptions for discoverability, progressive context reveal, and validation workflows. Critical for ensuring skills are discoverable and effective - prevents poorly structured skills that Claude won't use properly.
Use when user mentions reviewing PRs, provides GitHub PR URLs/numbers, or discusses code review. Provides structured analysis of code quality, backward compatibility, security issues, test coverage, and unaddressed comments with categorized findings (Critical/High/Medium/Low). Creates isolated git worktree for safe review, ensures comprehensive security analysis, and generates actionable recommendations. Invoke before analyzing any pull request changes.
Use when reviewing code or PRs in the llama-stack repository (llamastack/llama-stack on GitHub). Provides specialized knowledge about Llama Stack's architecture, testing patterns (recordings folders), backward compatibility requirements, distributed system considerations, and code review focus areas specific to this project.
Use when diagnosing openai_harmony.HarmonyError or gpt-oss tool calling issues with vLLM. Identifies error sources (vLLM server vs client), maps specific error messages to known GitHub issues, and provides configuration fixes for tool calling problems with gpt-oss models.
Use when you need to find the source code location of installed plugins, explore plugin structure, access reference materials in plugins, or debug plugin issues. Provides step-by-step guidance to trace installed plugins back to their local repository locations by reading ~/.claude/plugins/installed_plugins.json and navigating to the installPath. Useful for finding skills, commands, reference docs, or understanding plugin implementation.
Use when user asks where to put configuration, skills, or learnings, or discusses sharing config across projects. Provides decision criteria for the three-tier architecture (global ~/.claude, plugin, project-local .claude) to prevent duplication and ensure reusability. Invoke before creating new skills or configuration to determine the correct tier and location.
Use before creating or editing marketplace.json files, setting up marketplace repositories, or when user asks about plugin distribution. Provides expert guidance on marketplace structure, manifest configuration, plugin organization, semantic versioning, and git-based distribution workflows. Invoke when working with marketplace files or discussing how to share multiple plugins to ensure correct manifest structure and installation compatibility.
Use before creating or editing plugin.json manifests, organizing .claude-plugin directories, or when user asks about plugin structure, versioning, or distribution. Provides expert guidance on manifest configuration, component organization, semantic versioning, and shareable plugin best practices. Invoke when working with any plugin files or discussing plugin architecture to ensure correct structure and discoverability.