# 3.9G of regenerable renders — the readers rebuild these
output/*
# except the things no re-render can recreate: the user's marks and status
!output/notes/
output/notes/*.tmp
# …and by that same test, these are all recreatable, and they churn on every
# click or run: a .bak is yesterday's copy of the file beside it, and the
# rest are written by a harness command that can simply be run again
# (`quality`, `coverage`, `guard --review`). The marks, the file notes and
# the court status stay tracked — those are judgments, and nothing rebuilds
# them. So do the .log journals: they are the history the stale-mark test
# reads to say whether a rendering changed since it was marked.
output/notes/*.bak
# THE REVIEWER'S NOTES ARE THEIR OWN. filenotes.json and its journal carry
# what the reviewer wrote to themselves about a file, in their own words —
# not a product of this repo and not for it to publish (the user,
# 2026-08-24: 'i dont want my notes included'). The marks stay: a yay/nay is
# a verdict the guard reads, and it carries no prose.
output/notes/filenotes.json
output/notes/filenotes.log
output/notes/quality.json
output/notes/coverage.json
output/notes/guard-review.json
__pycache__/
*.pyc
.venv/

# Local rewrite/debugging artifacts
/after
/claudeplan.md
