MUST use before any commit operation. Detects VCS type (prefers jujutsu over git), enforces atomic commits, validates messages. Required before: 'git commit', 'jj commit', 'commit these changes', 'let me commit', 'ready to commit', 'git add', creating commits, branch operations, pull requests.
LLM-specific guidance for using jujutsu. Complements commit skill with jujutsu-specific workflows: working copy model, commit commands, file tracking, non-interactive operations. Triggers: jj commands, jujutsu workflows, working copy questions, bookmark management.
Produces and maintains a holistic system specification (SPEC.md) at a project root. The spec is the single source of truth for what the system must conform to. Use when starting a new project's spec, folding new ADRs or requirements into an existing spec, or reconciling contradicting decisions. Triggers: 'create SPEC.md', 'write a spec', 'system specification', 'update the spec', 'fold ADR into spec', 'specify the system'. Do NOT use for PRDs (run product-definition) or ADRs (run adr-writing); the spec is downstream of both.
Provides the md CLI for querying and mutating Markdown files. Prefer md over Edit for replacing or appending content between structural markers (headings, comments, nodes). Use when replacing sections, appending to sections, mutating frontmatter fields, extracting frontmatter, listing headings, finding links/wikilinks, counting words, extracting code blocks or comments, validating links, finding incoming links, format conversion (YAML/TOML), or batch processing .md files. Triggers: 'replace section', 'append content', 'edit markdown', 'update section', 'frontmatter', 'wikilinks', 'headings', 'markdown metadata', 'word count', 'broken links', 'code blocks', 'comments', 'footnotes', 'incoming links', 'backlinks', '.md files'. Do NOT use for full-text search across files (use grep/rg), non-markdown file formats, or rendering markdown to HTML.
Explores ideas through structured questioning to reach shared understanding. Use when brainstorming, requirements are unclear, or fleshing out any idea. Triggers: 'brainstorm', 'let us think about', 'flesh out this idea', 'explore options', unclear requirements, vague requests. Do NOT use for well-defined tasks with clear requirements.
Applies TDD methodology and research-backed practices (Meincke 2025) for writing or improving LLM prompts: measure baseline, test don't assume, iterate rigorously. Prevents assuming universal techniques work. Includes persuasion principles for compliance. Triggers: 'write a prompt', 'improve prompt', 'prompt not working', general prompting, application development. Not for formatting-only tasks requiring no iteration.
Creates and edits SKILL.md files for agent skills. Use when building or refining agent configuration. Enforces red-green-refactor methodology with agent-specific requirements: frontmatter format, 500-line limit, progressive disclosure, quality gates. Triggers: 'create a skill', 'write a skill', 'new SKILL.md', 'create agent documentation'. Do NOT use for general documentation or README files.
Use when implementing automated tests. Enforces empirical testing, test pyramid, TDD practices, concurrent testing. Triggers: writing tests, test planning, implementing test cases, TDD, race conditions. Go-specific patterns and rules: see references/go.md.