Create GitHub pull requests — picks the most token-efficient available method (gh CLI, GitHub MCP, or git push), and sources the PR body from an existing spec/plan instead of re-analyzing the diff
Prime — load project context by reading docs, the docs_context layer, core docs, and TODO before starting any task
Feature planning — research codebase, design options, write a structured spec/implementation plan to the project specs directory
Generate feature documentation by analyzing code changes and specifications — creates markdown docs in docs/core/ with conditional context
Bootstrap any project's AI workspace — detect the current tool, extract hardcoded paths from skill/command/agent files, write a central workspace.md, update all files to use config keys
Build or refresh the codebase graph indexes (codebase-memory-mcp structural index + graphify semantic graph) that other skills read from. The single writer — idempotent, incremental, safe to call repeatedly.
Group git changes by logical concern and generate a commit message for each group — follows Conventional Commits 1.0.0 and commitlint rules
Implement an approved spec — reads plan from specs_dir, executes phase by phase, updates checkboxes, verifies after each phase, and commits completed work