Writes React the way your team already agreed to write it.
Eval suite not run yet
Docs & writing
Docs That Match
Writes the guide by actually following it.
What it produces
step 1-4ran clean on a fresh image
step 5needs a dependency the docs never mention
step 6command renamed two releases ago
step 7-9ran clean
Trigger accuracy
Not measured yet. This skill does not go on sale until its eval suite has been run against it and the number published here.
What we changed
- Runs each step before writing it down, so the guide describes the software rather than the intention.
- Records the errors it hit, which become the troubleshooting section instead of being forgotten.
- Writes prerequisites from what was actually missing on a clean machine.
- Rechecks the guide after an edit, so it does not drift from the code the week after it is written.
- A troubleshooting table for projects whose setup cannot run unattended.
What it does, on real work
Writing an install guide
> write the setup docs for this project
Ran the setup on a clean image, documented the seven steps that worked and the two that needed a missing dependency.
Finding stale docs
> is the README still true?
Followed it end to end and reported three steps that no longer match the code.
Building a troubleshooting table
> what goes wrong for new people?
Listed the five errors hit during the clean run, each with the cause and the fix.
Requirements
- Claude Code 2.0+
- A container runtime for the clean-machine run
- macOS or Linux
Related
Cuts what Claude reads before it starts, without cutting what it knows.
Eval suite not run yet