Instinct-based learning system that observes sessions via hooks, creates atomic instincts with confidence scoring, and evolves them into skills/commands/agents.
Suggests manual context compaction at logical intervals to preserve context through task phases rather than arbitrary auto-compaction.
Use when asked to create diagrams, flowcharts, architecture visuals, or render Mermaid syntax to PNG/SVG files using the local mmdc CLI tool. Not for inline Mermaid in markdown.
Use when the user wants to review and optionally approve a pull request. Triggers: "review and approve PR", "approve PR #123", PR number/URL with approval intent, "LGTM this PR", "check and approve".
Generate a structured PR description from the current branch's changes. Use when the user says "generate PR description", "describe PR", "PR summary", "pr-describe", "describe my changes", or "write PR description".
Cache expensive file processing results using SHA-256 content hashes — path-independent, auto-invalidating, with service layer separation.
Decision framework for choosing between regex and LLM when parsing structured text — start with regex, add LLM only for low-confidence edge cases.
Scan your Claude Code configuration (.claude/ directory) for security vulnerabilities, misconfigurations, and injection risks using AgentShield. Checks CLAUDE.md, settings.json, MCP servers, hooks, and agent definitions.