.gitignore

_site/
_site-org/
_site-com/
.ssh/
dist/
tmp/
out-tsc/
build/
.nyc_output/

node_modules/

.idea/
.project
.classpath
.c9/
*.launch
.settings/
*.sublime-workspace
.vscode/*

.sass-cache/
connect.lock/
coverage/
typings/
/docs/

debug*
*.pid
*.log
.DS_Store
Thumbs.db
*.swp

package-lock.json
Gemfile.lock

.api-tmp/

# Private legal documents — kept off git (backed up privately). Never commit.
legal/

# Promotion working material — drafts, playbook and syndication exports.
# Local only: never commit this, the repo is public.
promotion/

# Internal planning notes. Local only, for the same reason as promotion/ above:
# this repo is public and carries the site, its tooling and its assets — not the
# working context of whoever built them.
#
# History, corrected on 2026-08-02 after checking GitHub's push record: the note
# above used to say the first two were "purged from history on 2026-07-31". No
# purge happened — master's push chain is unbroken, so nothing was ever
# force-pushed. They are absent from master because rebrand/dual-domain was
# integrated without becoming an ancestor of it, and they still exist in that
# branch's orphaned commits, which GitHub serves by SHA. CUTOVER.md and
# docs/superpowers/ were untracked on 2026-08-02 and are in master's history.
#
# The patterns come last so a new plan or report is ignored by default rather
# than by someone remembering to add it here.
#
# Two holes closed on 2026-08-05, both found the hard way: notes/mcp-search-docs-plan.md
# was committed and pushed to master. The patterns were UPPERCASE only, and gitignore is
# case-sensitive on Linux, so a lowercase `-plan.md` matched nothing; and a plan in a NEW
# directory escapes a root-anchored name pattern entirely. So: both cases, and the
# directory names a planning note is likely to be filed under.
BLOG-BACKLOG.md
REBRAND-PLAN.md
API-DOCS-PLAN.md
PROOFREAD-REPORT.md
CUTOVER.md
*-PLAN.md
*-NOTES.md
*-REPORT.md
*-HANDOVER.md
*-plan.md
*-notes.md
*-report.md
*-handover.md
notes/
plans/

# Build intermediate: the curated front matter the search index folds in. Moved here out
# of the output dir by scripts/lib/search-corpus.js so an incremental rebuild can still
# read it — see takeFrontmatter() for why deleting it was a correctness bug.
.search-frontmatter-*.json

# The artificial KPI query set: 1.2 MB, and regenerated exactly by
# `npm run kpi:search:gen` from the built index plus a fixed PRNG seed, so committing it
# would only duplicate the content it is derived from. The NATURAL set beside it IS
# committed, because a Google autocomplete harvest cannot be reproduced — the suggestions
# change — and without the same queries a change in the KPI would not mean a change in the
# ranker. See scripts/search-kpi/README.md.
scripts/search-kpi/data/artificial-queries.json