ssamba1

Verify that a repository's README quickstart actually runs, and capture a real demo of the project working. Use when the user says "/proof", asks to verify a README, check that a quickstart works, prove install/usage commands run, or generate a real demo GIF/screenshot for a repo. Routes to the proof-verify and proof-demo skills.

Capture a real demo of a project running — a terminal GIF for a CLI (via vhs) or a screenshot for a web app (via Playwright), with a real text-transcript fallback. Never fabricates or AI-generates the image. Use when asked to make a demo, record the CLI, screenshot the app, or run "/proof demo".

Run a repository's README quickstart in a sandbox and report whether it actually works, correcting documentation errors but never masking real code bugs. Use when asked to verify a README, check a quickstart, prove install/usage commands run, or run "/proof verify".

Humanize AI-generated text via a closed-loop, detector-feedback rewrite (the "untell" skill). Use when the user wants to humanize text, make writing sound more human or less like AI, reduce AI-detection scores, bypass or beat an AI detector (GPTZero, Turnitin, Originality, ZeroGPT), or lower the "AI probability" of a passage or file. Research/defensive tool — preserves meaning, citations, and facts.

Audit your INSTALLED Claude Code skill library for waste. Measures the per-turn token tax of every skill's always-on description, mines transcripts to find skills that never fire, detects trigger collisions between skills, flags stale model references, and emits the exact disable-model-invocation edits to cut the tax. Use when Claude Code feels bloated or slow, when you've installed many skills/plugins, to reduce context cost, prune unused skills, find duplicate or conflicting skills, or audit your ~/.claude/skills library.