# Rust build output
/target
/target-seam
**/*.rs.bk

# Harness / SPT local state (machine-specific)
.claude/

# OS junk
Thumbs.db
.DS_Store
.gstack/

# mdBook build output (docs-site truth is src/; rendered book publishes via CI)
docs-site/book/

# Gate/agent worktrees — ALL git worktrees live here (doyle convention 2026-07-05);
# never as siblings of the project dir. Each carries its own throwaway target.
.worktrees/

# Build-step benchmark ledger (releases#108). NEVER committed: a hand-maintained
# ledger file in git is the conflict-magnet class every lane would touch and every
# merge would fight over. CI emits its own per-run artifact instead.
.bench-ledger/
# …and the CI wraps' default relative directory, which lands in the checkout itself.
# Ignored for the same reason plus one more: an untracked directory in the workspace
# reads as a DIRTY TREE to anything that asks git (debug-rollout stamps `git_dirty`
# into an update set's provenance from `git status --porcelain`).
bench-ledger/

# Agent scratch. Peer messages, commit/PR bodies, gate legs, run output and preserved
# evidence — 152 entries / 277 MiB on the authoring box the day this landed, none of it
# repo content, all of it staged by a `git add -A` at the root because nothing ignored it.
#
# SIX FILES MOVED OUT rather than behind this line, because the repo cites them:
#   .spt/rig/twohost-web-local.sh            -> .github/bench/twohost-web-local.sh
#   .spt/preserved/pump-on-one-box-rig.patch -> docs/spikes/pump-on-one-box-rig.patch
#   .spt/RCA-FLEET-DAEMON-14444.md           -> docs/RCA-FLEET-DAEMON-14444.md
#   .spt/PID-KILL-CENSUS.md                  -> docs/PID-KILL-CENSUS.md
#   .spt/launch-battery.py                   -> .github/bench/launch-battery.py
#   .spt/RIDER-65S-DESIGN.md                 -> docs/design/RIDER-65S-DESIGN.md
# launch-battery.py is the reason this list exists: its own docstring records that the
# original was LOST to a `.spt/` sweep in 2026-09 and rebuilt at cost, while the register
# cites it. An ignore line is the condition that produced that loss, so anything the repo
# names moves out before the line goes in.
#
# Any `.spt/...` path cited from docs/ is MACHINE-BOUND EVIDENCE on the box that produced
# it, never a repo path; this line neither tracks nor deletes it. Six such citations at
# landing, none of them repo content. Deliberately a RULE and a COUNT, not a list: the
# register gains evidence citations continuously (two arrived the same day this landed),
# and an enumeration in an ignore comment is stale the next time anyone files a finding.
.spt/

# Gate legs written at the REPO ROOT by battery/driver scripts. Root-anchored on
# purpose (`/*.raw`, not `*.raw`): these patterns must not reach into crates/ or
# docs/, where a `.raw` fixture would be repo content.
#
# 473 such files existed when this landed (268 .raw, 184 .exit, 7 .out, 4 .done,
# 4 .log, 2 .stackdump, 2 .sh, 2 .txt — 45 MiB). Every one was checked for a
# tracked citation first, word-bounded against the whole tracked corpus, and had
# none; they were MOVED to .spt/preserved/root-gate-legs-2026-09-09/ with a
# sha256 manifest, not deleted.
#
# Legs are the one class whose FUTURE instances are scratch by construction, so
# ignoring them institutionalizes no loss — unlike a directory ignore, which is
# how .spt/launch-battery.py was lost once already. A leg kept as evidence lives
# under .spt/preserved/ and is cited by THAT path, never by a root path.
#
# Deliberately NOT ignored: `/*.md`, `/*.log`, `/*.sh`, `/*.txt`. A plan or a
# script dropped at the root MUST keep showing as `??` so the next census
# classifies it — that is the pattern which produced this repo's tracked root
# corpus, and burying it would hide the next one.
/*.raw
/*.exit
/*.done
/*.out
/*.stackdump