readme-updaterlisted
Install: claude install-skill mistakenot/auto-stack
# README Updater
Bring a stale README back into sync with reality, end to end, in one pass.
The skill runs four phases sequentially without pausing for human input,
then commits and pushes the result.
## When this skill triggers
- "update the README", "the README is out of date", "refresh README.md"
- "the README doesn't mention <feature>"
- "rewrite the README to cover what we've shipped"
- Any request to bring docs in line with the current state of the code
It does **not** trigger for:
- Small spot-edits to a README (just edit directly)
- Drafting a brand-new README from scratch in an empty repo (different problem)
- Generating per-package docs (use a docgen tool)
## The most important lesson
**Never trust your memory or the existing README for CLI claims.**
On every prior run, the model has shipped at least one snippet referencing
a flag that no longer exists or a subcommand that was renamed. The
verification phase below is non-negotiable — it is the one step that
distinguishes a useful README update from a confidence-damaging one.
## Phase 1 — Scope
Find when the README was last meaningfully updated and what's changed
since. This bounds the work.
```bash
# Last commits that touched README.md (not just chore renames)
git log --oneline -20 README.md
# Pick the most recent substantive update commit — call it $LAST_README_SHA
git log --oneline $LAST_README_SHA..HEAD | wc -l # how many commits since
git log --oneline $LAST_README_SHA..HEAD # the actual li