Post-process prose to strip AI-telltale patterns — puffery, bloat, meta-commentary, false authority, and editorial framing.
Create a git commit with trailers for session cost metrics and Claude attribution. Use when the user asks you to commit changes to git.
Check Claude Code changelog for updates since your last commit. Use when the user wants to see what changed in Claude Code.
Runs the project test suite, analyzes failures, and proposes fixes via sub-agents.
Display current session cost metrics. Use when the user asks to see current costs, cost usage, or how much the session has cost so far.
Experimental. Aggressively distill LLM-facing prose to imperative essentials. Skills → 1–3 sentences. Procedural skills → numbered list. Multi-section instruction files → bullets per section.
Tighten prose for LLM consumption. Strip filler, switch to imperative voice, collapse verbose lists. Apply to any markdown documentation — skill files, agents, references, rules, CLAUDE.md
Inspect commits since last release and compose changelog entries. Use when the user wants to update their changelog.