BFM-decomp/.gitignore

# ============================================================
# ROM-content policy — PROJECT_CONTEXT.md rule H1, IN FORCE (Phase 33, the public flip).
#
# The repository ships NO ROM-derived bytes and NO proprietary SDK: the game disc, the
# extracted payloads (`extracted/`, regenerated from YOUR dump by `make disc-extract` and
# verified against the committed manifest), the splat disassembly (`asm/`, `assets/`), the
# RAM-dump corpus (`dumps/*.bin`), the Ghidra project (`ghidra/`, whose database embeds the
# program bytes — regenerable from text via tools/ghidra_rebuild.sh), Sony's PsyQ SDK
# (`tools/psyq/`, optional and user-supplied), the session-transcript archive, and
# redistributed third-party binaries. Symbol names, addresses, hashes, configs and the
# decompiled C are what a decomp publishes (sotn precedent). The private history that
# carried these while the repo was private (rule R1, 2026-06-10 .. 2026-09-06) lives in the
# archive repo only.
# ============================================================

# Raw disc dump — never committed (size + reproducible; your own dump goes here)
/Brave Fencer Musashi (USA)/
/disks/

# RAM-dump corpus — 28 × 2 MB images of the running game (local-only; dumps/INDEX.md +
# dumps/CHECKSUMS.sha1 stay tracked and describe them)
/dumps/*.bin

# The Claude Code session-transcript archive (private; the archive repo holds it)
/session archive/

# Redistributed third-party binaries: download per docs/SETUP.md (sha256 recorded there)
/tools/ghidra-ext/*.zip
/tools/brave-CUE/brave.exe

# Project-local runtime scratch (MCP server log + stop sentinel) — replaces /tmp use.
# Contents-exclude form (the /tools/bin/*.sha256 precedent) so the IRREPLACEABLE recon
# commits while the regenerable bulk stays ignored (R20 backup vs R12 scratch; Phase-27 T3).
# Rule of thumb: commit what a rerun CANNOT reproduce (hand/Fable5 analysis, harnesses,
# ledgers); leave anything a script regenerates (RTL dumps, build logs, drafts) ignored.
/.run/*

# --- Preserved recon (R20) ---
# Giant crack recon: the Phase-25 *.opus.{c,md} seeds, the byte-verified pf*.c
# regression ladder, the dump/monitor harnesses, and — cited BY cookbook §45 as its
# worked example — func_80133CD4.fable.{c,md} + the fable_cd4/ gdb oracle.
# The d_pf*.i.* / *.s / dumps_*/ gcc RTL scratch beside them is regenerable via
# dump.sh and stays ignored, as do the permuter/ILS .log files.
!/.run/giants/
/.run/giants/*
!/.run/giants/*.c
!/.run/giants/*.md
!/.run/giants/*.sh
!/.run/giants/*.py
# P30 S42 serial crack drafts — ~680k subagent tokens of work, both NEAR and both the best
# base any future attempt has. They were gitignored, i.e. one `git clean` from gone (R20).
!/.run/s42/
/.run/s42/*
!/.run/s42/*/
!/.run/s42/*/*.c
# P30 S43 — the permuter-ILS improvement on the same function (63 -> 42 mismatched), the best
# base that exists for it. Same R20 reasoning as s42 above.
!/.run/s43/
/.run/s43/*
!/.run/s43/*.c
!/.run/s43/*.log
!/.run/s43/*.py
!/.run/s43/*.json
!/.run/probe_jtbl/
/.run/probe_jtbl/*
!/.run/probe_jtbl/verdict.md
!/.run/probe_jtbl/rodata_verbatim.objdump
!/.run/probe_jtbl/rodata_filtered.objdump
!/.run/giants/fable_cd4/
/.run/giants/fable_cd4/*
!/.run/giants/fable_cd4/*.c
!/.run/giants/fable_cd4/*.md
!/.run/giants/fable_cd4/*.sh
!/.run/giants/fable_cd4/*.gdb
!/.run/giants/fable_cd4/*.txt
# P33.5 task 8: the two full disassembly listings of func_80133CD4 are ROM-derived (class 3 of the firewall) and were
# untracked; they stay in the published history by owner decision (tools/public_rewrite/untracked_after_rewrite.txt).
/.run/giants/fable_cd4/*_full.txt

# Phase-26 gdb-on-cc1 oracle harness + proof + draft ladder for func_80178004 (the
# distilled output of a 477k-token Fable5 pass; the method §52 lever 6 depends on).
# dumps_*/ are regenerable via runorc.sh + the .gdb scripts -> stay ignored.
!/.run/fable_80178004/
/.run/fable_80178004/*
!/.run/fable_80178004/*.c
!/.run/fable_80178004/*.md
!/.run/fable_80178004/*.gdb
!/.run/fable_80178004/*.sh
!/.run/fable_80178004/*.py
!/.run/fable_80178004/*.lst

# The frontier ledgers every survey, worklist and plan consumes.
!/.run/backlog.jsonl
# S79/S80 task #9 — one-agent-per-function drafting: the arm drafts (the NEAR seeds), the briefs' laws, the
# target list and the verdict ledger (the PhaseEnd's wall-proof source). Scratch under opus/w_* etc. stays ignored.
!/.run/S79w/
/.run/S79w/*
!/.run/S79w/SYS.md
!/.run/S79w/targets.json
!/.run/S79w/verdicts/
/.run/S79w/verdicts/*
!/.run/S79w/verdicts/*.jsonl
!/.run/S79w/haiku/
/.run/S79w/haiku/*
!/.run/S79w/haiku/func_*.c
!/.run/S79w/sonnet/
/.run/S79w/sonnet/*
!/.run/S79w/sonnet/func_*.c
!/.run/S79w/opus/
/.run/S79w/opus/*
!/.run/S79w/opus/func_*.c
!/.run/S79w/permuter/
/.run/S79w/permuter/*
!/.run/S79w/permuter/func_*.c
!/.run/fuel_manifest.json

# Phase-27 pin-crash characterization: the minimal repro + gdb harness (the irreplaceable ~KB
# core; pin_crash_sigabrt.md above is already covered by giants/*.md). The 260M of RTL dumps +
# bisection .s/.c beside them are regenerable via abort.gdb -> stay ignored (R33).
!/.run/giants/pincrash/
/.run/giants/pincrash/*
!/.run/giants/pincrash/minimal_repro.c
!/.run/giants/pincrash/*.gdb
!/.run/giants/pincrash/*.md

# Regenerated build output (deterministic churn)
/build/
/expected/

# Regenerated splat outputs — recreated by `make extract` from the committed
# config/splat.*.yaml + config/symbols*.txt + the extracted payloads. The durable RE record
# is the symbol file, not these (build scaffolding, not committed work). The four splat preset
# headers (include/include_asm.h, macro.inc, labels.inc, gte_macros.inc) are TRACKED since P33 B1:
# generic presets, identical for every binary, ROM-free — the compile-only CI needs them.
/asm/
/assets/
/undefined_syms_auto.txt
/undefined_funcs_auto.txt

# Ghidra project — its database embeds the program bytes (ROM-derived), so it is not tracked
# (P33 B1; it was committed while private under rule R20). The RE work it holds is exported to
# text (config/ghidra/*.jsonl) and the project is rebuilt from that + the disc by
# tools/ghidra_rebuild.sh. NEVER `git clean -x` here — that deletes the working database.
/ghidra/

# Toolchain downloads & caches (not source)
/.venv/
__pycache__/
# Compiler binaries are ignored (regenerable downloads); the committed *.sha256
# checksum record is re-included so a fresh clone can verify its download (sotn
# pattern). Contents-exclude form (not /tools/bin/) so the !re-include can apply.
/tools/bin/*
!/tools/bin/*.sha256
# PsyQ SDK — Sony-copyrighted; OPTIONAL and user-supplied (tools/fetch_psyq.sh), never distributed.
# The whole directory is ignored (P33 B1; the small working artifacts were committed while private
# under rule R20 and live in the archive repo). The build is byte-identical without it (the
# INCLUDE_ASM fallback tiles); with it, main links Sony's real objects (`make sdk-dual` proves both).
# The checksums of the archives a user must supply are tracked at tools/psyq_CHECKSUMS.sha256.
/tools/psyq/
/tools/pcsx-redux/
*.tar.gz
# ...EXCEPT the old-gcc cc1 compiler tarballs — back them up (gathered compilers; small; verified by
# tools/bin/CHECKSUMS.sha256). Extracted cc1 binaries stay ignored (regenerated by `tar xzf`, SETUP §4.7). R20.
!/tools/bin/*.tar.gz

# Cross-validation scratch: compiled brave oracle + its ROM-derived .dir output
/tools/brave-CUE/_build/

# Phase-18 compiler-quirk research reference repos (gitignored; re-clonable from GitHub, exact SHA
# recorded in docs/SETUP.md). pmret/gcc-papermario = the gcc-2.7.2 source (read the reload/CSE/
# scheduling/regalloc passes) + ladysilverberg/xenogears-decomp = our exact compiler's decomp (mine
# transferable idioms). Research reference, not a build dependency — not committed (avoids repo bloat).
/tools/reference/

# Editor / OS noise
.vscode/
Thumbs.db

# Asset/EXE extraction: commit the tooling + manifest + the retail EXE, NOT the
# bulk payloads (deterministically regenerable from the dump; ~hundreds of MB
# incl. FMV) and NOT the prototype EXEs. Per-ROM subfolders under extracted/:
#   extracted/retail/  — retail SLUS-00726 disc extraction (3 files committed)
#   extracted/proto/   — prototype main EXEs (ROM-derived; fully ignored)
# Nested allowlist: ignore everything under extracted/, re-include only the 2 manifest
# files (hashes, never bytes — the oracle `make disc-extract` verifies YOUR extraction
# against). The EXE itself left the tree at P33 B1 (H1): it is regenerated from your disc.
/extracted/*
!/extracted/retail/
/extracted/retail/*
!/extracted/retail/manifest.jsonl
!/extracted/retail/manifest.sha1

# NOTE: asm/ + assets/ + undefined_*_auto.txt are IGNORED above: `make extract` regenerates
# them deterministically from the extracted payloads
# from the committed yaml + symbols, so they are build scaffolding, not durable RE work.
# (Phase 5 refined the earlier "commit asm/" stance once the regeneration flow existed;
# the durable record is config/symbols.us.txt.)

# Gen2 matching-model corpus + weights (R20: large artifacts out of git)
datasets/
models/
*.gguf
*.safetensors
.venv-train/
unsloth_compiled_cache/
.venv-serve/
.env

# Phase 29 SESSION-22 — the wave22 UNBANKED drafts. 13 of the 18 (6 NEAR with precise residual
# diagnoses + 7 that reached match_one MATCH but did not bank whole-binary) are the next session's
# named fuel and cost ~2.59M subagent tokens. They lived only on disk, i.e. one `git clean -fdx`
# from gone — the same exposure the .run/giants Fable5 cracks had before Phase 27 curated them.
!/.run/wave22/
/.run/wave22/*
!/.run/wave22/*.c
!/.run/wave22_targets.json

# Phase 29 SESSION-23 — the NEAR-6 measurement corpus + the improved ILS seeds. The two
# permuter waypoints (func_80177B5C 11->7, func_80140958 10->6) are CPU-bought and second-oracle
# verified; the f132F40_v*.c ladder is the byte-evidence for a NEGATIVE result (the §2 remat
# recipe does not dissolve this hoist), which is exactly the kind of finding that gets re-derived
# expensively if it is lost. near6_measure.{py,json} is the reproducer.
!/.run/near6/
/.run/near6/*
!/.run/near6/*.c
!/.run/near6_measure.py
!/.run/near6_measure.json
!/.run/near6_ils_*.log
# P30 S45 — the L3 tour's irreplaceable evidence (R20): the state-change tour log + hit records,
# the watcher/derivation scripts, and the module rosters. The 104 ram_*.bin dumps (~208 MB) stay
# LOCAL-ONLY: regenerable via another debug-menu tour; the byte-proofs live in the jsonl + docs.
!/.run/s45/
/.run/s45/*
!/.run/s45/*.py
!/.run/s45/*.sh
!/.run/s45/*.jsonl
!/.run/s45/*.tsv
!/.run/s45/onboard*.log

# P32 T3 (2026-09-05): the one-agent-per-function wave's ledger + drafts (R20 — the drafts are the only copy of
# byte-verified MATCH/NEAR bodies outside the harness transcripts; scratch dirs stay ignored)
!/.run/P32/
/.run/P32/*
!/.run/P32/*.json
!/.run/P32/t3/
/.run/P32/t3/*
!/.run/P32/t3/SYS.md
!/.run/P32/t3/BRIEF.md
!/.run/P32/t3/LAUNCHED.md
!/.run/P32/t3/targets.json
!/.run/P32/t3/verdicts.jsonl
!/.run/P32/t3/pending_launch.txt
!/.run/P32/t3/harvest_notes.md
!/.run/P32/t3/PROMPT_TEMPLATE.md
!/.run/P32/t3/reports/
/.run/P32/t3/reports/*
!/.run/P32/t3/reports/*.md
!/.run/P32/t3/haiku/
/.run/P32/t3/haiku/*
!/.run/P32/t3/haiku/func_*.c
!/.run/P32/t3/sonnet/
/.run/P32/t3/sonnet/*
!/.run/P32/t3/sonnet/func_*.c
!/.run/P32/t3/opus/
/.run/P32/t3/opus/*
!/.run/P32/t3/opus/func_*.c
!/.run/P32/t3/restored/
/.run/P32/t3/restored/*
!/.run/P32/t3/restored/sweep_*.py
# P32 T3 session 491895ad (S83): the resume session's helpers, prompts, gate slates and logs (not the verify/ compile dirs)
!/.run/P32/t3s3/
/.run/P32/t3s3/*
!/.run/P32/t3s3/*.sh
!/.run/P32/t3s3/*.py
# P33.5 task 8: of the session's 41 root logs only the three the PhaseEnd cites stay tracked (the R22 runs); the
# bank_*/report/tools-health logs of a finished session are regenerable evidence and were untracked (still on disk)
!/.run/P32/t3s3/r22_check.log
!/.run/P32/t3s3/r22_full.log
!/.run/P32/t3s3/r22b_full.log
!/.run/P32/t3s3/*.txt
!/.run/P32/t3s3/prompts/
!/.run/P32/t3s3/gate/
/.run/P32/t3s3/gate/*
!/.run/P32/t3s3/gate/*.json
!/.run/P32/t3s3/gate/*.log
# P32 T4 (S83): the walls' re-probe artefacts — synced/variant drafts only (the sandbox TU copies and rtu compile dirs are regenerable)
!/.run/P32/t4/
/.run/P32/t4/*
!/.run/P32/t4/drafts/
!/.run/P32/t4/v32A74/
/.run/P32/t4/v32A74/*
!/.run/P32/t4/v32A74/*.c
!/.run/P32/t4/v391D4/
/.run/P32/t4/v391D4/*
!/.run/P32/t4/v391D4/*.c
# P32 T4b/T5x (S83): the hand-pass notes + measured variant drafts, and the Fable pass's brief/targets/drafts/reports
!/.run/P32/t4b/
/.run/P32/t4b/*
!/.run/P32/t4b/func_*/
/.run/P32/t4b/func_*/*
!/.run/P32/t4b/func_*/*.md
!/.run/P32/t4b/func_*/*.c
!/.run/P32/t4b/*.log
# P33.5 task 8: the uncited duplicate close-report logs of this step were untracked (the cited copies live under t4d/t4e)
/.run/P32/t4b/report_close.log
/.run/P32/t4b/report_main_close.log
!/.run/P32/t5x/
/.run/P32/t5x/*
!/.run/P32/t5x/BRIEF.md
!/.run/P32/t5x/targets.json
!/.run/P32/t5x/fable/
/.run/P32/t5x/fable/*
!/.run/P32/t5x/fable/func_*.c
!/.run/P32/t5x/reports/
/.run/P32/t5x/reports/*
!/.run/P32/t5x/reports/*.md
!/.run/P32/t5x/verdicts.jsonl
!/.run/P32/t5x/resume_queue.txt
# P32 T4b hand pass (S84, 2026-09-06): notes, mechanism probes, verdict drafts and the private dump script (not the dumps_*/ or rtu/ dirs)
!/.run/P32/t4c/
/.run/P32/t4c/*
!/.run/P32/t4c/*.sh
!/.run/P32/t4c/*.py
!/.run/P32/t4c/*.md
!/.run/P32/t4c/*.log
!/.run/P32/t4c/*.json
!/.run/P32/t4c/func_*/
/.run/P32/t4c/func_*/*
!/.run/P32/t4c/func_*/*.c
!/.run/P32/t4c/func_*/*.md
# P32 T4b hand pass row (d) (S85, 2026-09-06): the sibling-port variants, their rtu verdict logs, the notes and the banked body (not the rtu/ compile dirs)
!/.run/P32/t4d/
/.run/P32/t4d/*
!/.run/P32/t4d/*.c
!/.run/P32/t4d/*.md
!/.run/P32/t4d/*.log
# P33.5 task 8: uncited duplicates untracked (report_close.log and tools_health.log here ARE cited and stay)
/.run/P32/t4d/audit_disc.log
/.run/P32/t4d/report_main_close.log
!/.run/P32/t4d/bank/
/.run/P32/t4d/bank/*
!/.run/P32/t4d/bank/*.c
# P32 T4c hand pass (S85, 2026-09-06): the two main rows' variants, rtu logs, notes, reproducers, bank bodies, slates, gate/R22 logs (not dumps_*/ or rtu/)
# P33 A5 (S87, 2026-09-06): THE recorded contract run — tools/verify_contract.sh's per-step logs, the two link
# maps of the with/without-SDK dual and SUMMARY.md (quoted by docs/verification.md). Evidence, tracked.
!/.run/P33/
/.run/P33/*
!/.run/P33/verify/
/.run/P33/verify/*
!/.run/P33/verify/*.log
# P33 E1 (S89, 2026-09-07): the decomp.me toolchain-replica verdict log (tools/decompme_replica.sh) — the log only; the
# regenerated target .s (game disassembly, H1), the probe objects and the old-gcc tarball under .run/P33/e1/ stay ignored
!/.run/P33/e1/
/.run/P33/e1/*
!/.run/P33/e1/decompme_replica.log
!/.run/P33/verify/*.map
!/.run/P33/verify/*.md
!/.run/P33/verify/*.txt
!/.run/P32/t4e/
/.run/P32/t4e/*
!/.run/P32/t4e/*.c
!/.run/P32/t4e/*.md
!/.run/P32/t4e/*.log
# P33.5 task 8: the uncited duplicate untracked (r22_*.log, report_close.log and tools_health.log here are cited and stay)
/.run/P32/t4e/report_main_close.log
!/.run/P32/t4e/*.json
!/.run/P32/t4e/*.sh
!/.run/P32/t4e/*.py
!/.run/P32/t4e/bank/
/.run/P32/t4e/bank/*
!/.run/P32/t4e/bank/*.c
!/.run/P32/t4e/ghost/
/.run/P32/t4e/ghost/*
!/.run/P32/t4e/ghost/*.c

# P33.5 task 13 (S91, 2026-09-07): the decomp-architect dry-run fixture — the answers file, the expected manifest, the
# before/after assertions and the install logs are evidence, tracked; the nested throwaway repo/ stays ignored.
!/.run/P33.5/
/.run/P33.5/*
!/.run/P33.5/kit-dryrun/
/.run/P33.5/kit-dryrun/*
!/.run/P33.5/kit-dryrun/*.md
!/.run/P33.5/kit-dryrun/*.txt
!/.run/P33.5/kit-dryrun/*.log
!/.run/P33.5/kit-dryrun/*.json
!/.run/P33.5/kit-dryrun/*.py

# P33.5 task 14.5 (S92, 2026-09-07): the phase-log mining pass — the brief, the 21 slice reports, the harvest and its table,
# the two scripts. Evidence of what the kit's kernels DK-69–DK-80 were distilled from; quotes of our own worklogs, nothing ROM-derived.
!/.run/P33.5/log-mining/
/.run/P33.5/log-mining/*
!/.run/P33.5/log-mining/*.md
!/.run/P33.5/log-mining/*.py

# P33.5 task 13.5 (S91, 2026-09-07): the tools-audit evidence — the survey inputs, the review verdicts, the need-keys, the layout
# probe and the run-4 judgment. Evidence, tracked.
!/.run/P33.5/tools_audit/
/.run/P33.5/tools_audit/*
!/.run/P33.5/tools_audit/*.txt
!/.run/P33.5/tools_audit/*.tsv
!/.run/P33.5/tools_audit/*.json

# P34 (S93, 2026-09-08): the flip phase's evidence — the probe logs (before/after the flip), the GitHub API and curl outputs of
# the post-flip checks, the outward-action records, the C11 prune listing, the close tools-health log. Text only, nothing ROM-derived.
!/.run/P34/
/.run/P34/*
!/.run/P34/*.log
!/.run/P34/probe/
/.run/P34/probe/*
!/.run/P34/probe/*.log
!/.run/P34/flip/
/.run/P34/flip/*
!/.run/P34/flip/*.log
!/.run/P34/flip/*.json
!/.run/P34/flip/*.txt
!/.run/P34/outward/
/.run/P34/outward/*
!/.run/P34/outward/*.log
!/.run/P34/outward/*.txt
!/.run/P34/outward/*.md
!/.run/P34/c11/
/.run/P34/c11/*
!/.run/P34/c11/*.log
!/.run/P34/c11/*.txt

# P35 (S94, 2026-09-08): the dedup phase's evidence — the R22 baseline and close logs, the probe's object A/B ledger and
# measurements, the census JSON/tables, the share runs' logs and ledgers. Text only, nothing ROM-derived (no .o, no .s, no images).
!/.run/P35/
/.run/P35/*
!/.run/P35/*.log
!/.run/P35/*.md
!/.run/P35/baseline/
/.run/P35/baseline/*
!/.run/P35/baseline/*.log
!/.run/P35/baseline/*.txt
!/.run/P35/probe/
/.run/P35/probe/*
!/.run/P35/probe/*.log
!/.run/P35/probe/*.txt
!/.run/P35/probe/*.md
!/.run/P35/probe/*.py
!/.run/P35/census/
/.run/P35/census/*
!/.run/P35/census/*.json
!/.run/P35/census/*.txt
!/.run/P35/census/*.md
!/.run/P35/share/
/.run/P35/share/*
!/.run/P35/share/*.log
!/.run/P35/share/*.json
!/.run/P35/share/*.txt
!/.run/P35/convert/
/.run/P35/convert/*
!/.run/P35/convert/*.log
!/.run/P35/convert/*.json
!/.run/P35/convert/*.txt
!/.run/P35/convert/*.sh

# P36 (S97, 2026-09-09): the levers-off phase's evidence — the R22 baseline and close logs, the lever census JSON/tables, the
# oracle calibration and the probe's measurements, the delever ledger and batch logs, the permuter outcomes, the wave packs and
# verdicts. Text only, nothing ROM-derived (no .o, no .s, no images). Contents form (R84).
!/.run/P36/
/.run/P36/*
!/.run/P36/*.log
!/.run/P36/*.md
!/.run/P36/baseline/
/.run/P36/baseline/*
!/.run/P36/baseline/*.log
!/.run/P36/baseline/*.txt
!/.run/P36/census/
/.run/P36/census/*
!/.run/P36/census/*.json
!/.run/P36/census/*.txt
!/.run/P36/census/*.md
!/.run/P36/probe/
/.run/P36/probe/*
!/.run/P36/probe/*.log
!/.run/P36/probe/*.txt
!/.run/P36/probe/*.md
!/.run/P36/probe/*.json
!/.run/P36/probe/*.py
!/.run/P36/probe/*.jsonl
!/.run/P36/delever/
/.run/P36/delever/*
!/.run/P36/delever/*.log
!/.run/P36/delever/*.json
/.run/P36/delever/recipes.json
/.run/P36/delever/inflight.json
/.run/P36/delever/includers_cache.json
!/.run/P36/delever/*.jsonl
!/.run/P36/delever/*.txt
!/.run/P36/gte/
/.run/P36/gte/*
!/.run/P36/gte/*.json
!/.run/P36/gte/*.log
!/.run/P36/gte/*.txt
/.run/P36/gte/inventory.json
!/.run/P36/engine/
/.run/P36/engine/*
!/.run/P36/engine/*.log
!/.run/P36/engine/*.md
!/.run/P36/engine/*.json
!/.run/P36/engine/*.jsonl
!/.run/P36/engine/*.txt
!/.run/P36/engine/trace/
/.run/P36/engine/trace/*
!/.run/P36/engine/trace/*.jsonl
!/.run/P36/agents/
/.run/P36/agents/*
!/.run/P36/agents/*.md
!/.run/P36/agents/*.tsv
!/.run/P36/agents/*__func_*/
/.run/P36/agents/*__func_*/*
!/.run/P36/agents/*__func_*/*.txt
!/.run/P36/agents/*__func_*/*.md
!/.run/P36/agents/*__func_*/body*.c
!/.run/P36/permuter/
/.run/P36/permuter/*
!/.run/P36/permuter/*.log
!/.run/P36/permuter/*.json
!/.run/P36/permuter/*.jsonl
!/.run/P36/permuter/*.txt
!/.run/P36/waves/
/.run/P36/waves/*
!/.run/P36/waves/*.log
!/.run/P36/waves/*.json
!/.run/P36/waves/*.jsonl
!/.run/P36/waves/*.txt
!/.run/P36/waves/*.md

# P37 (S106, 2026-09-12): the structs phase's evidence — the R22 baseline, oracle calibration and close logs, the type census
# JSON/tables (+ parked.tsv), the three probes' tables, the restruct ledger and batch logs, the per-TU declaration judgements, the
# agent packs and verdicts. Text only, nothing ROM-derived (no .o, no .s, no images). Contents form (R84).
!/.run/P37/
/.run/P37/*
!/.run/P37/*.log
!/.run/P37/*.md
!/.run/P37/baseline/
/.run/P37/baseline/*
!/.run/P37/baseline/*.log
!/.run/P37/baseline/*.txt
!/.run/P37/census/
/.run/P37/census/*
!/.run/P37/census/*.json
/.run/P37/census/struct_map.json
/.run/P37/census/body_base_type.json
!/.run/P37/census/*.txt
!/.run/P37/census/*.md
!/.run/P37/census/*.tsv
!/.run/P37/probe/
/.run/P37/probe/*
!/.run/P37/probe/*.log
!/.run/P37/probe/*.txt
!/.run/P37/probe/*.md
!/.run/P37/probe/*.json
!/.run/P37/probe/*.jsonl
!/.run/P37/probe/*.py
!/.run/P37/restruct/
/.run/P37/restruct/*
!/.run/P37/restruct/*.log
!/.run/P37/restruct/*.json
# the argcheck rows cached on the src stamp — regenerable (a trailing comment on a pattern line is PART of the pattern: own line)
/.run/P37/restruct/argcheck_cache.json
/.run/P37/restruct/inflight.json
/.run/P37/restruct/includers_cache.json
!/.run/P37/restruct/*.jsonl
!/.run/P37/restruct/*.txt
!/.run/P37/decls/
/.run/P37/decls/*
!/.run/P37/decls/*.log
!/.run/P37/decls/*.json
!/.run/P37/decls/*.jsonl
!/.run/P37/decls/*.txt
!/.run/P37/agents/
/.run/P37/agents/*
!/.run/P37/agents/*.md
!/.run/P37/agents/*.tsv
!/.run/P37/agents/*__func_*/
/.run/P37/agents/*__func_*/*
!/.run/P37/agents/*__func_*/*.txt
!/.run/P37/agents/*__func_*/*.md
!/.run/P37/agents/*__func_*/body*.c

# Outreach drafts (P34 task 6, Drew's decision 2026-09-08): the notes to neighbouring projects were SENT (the Archipelago issue,
# the decomp.me preset request, the tools post, the #ai reply); the drafts stay on the maintainer's machine and leave the tree
# from here (no rewrite — they are in the published history). Contents form, R84.
/docs/outreach/*

# The document archive (P34 task 7, Drew's decision 2026-09-08): docs/sunset/ (59 documents retired at Phase 33.5) leaves the tree —
# the wiki's Archive index rows are the record, the last tracked versions are in the history (v1.32.1). A local copy may remain;
# ignored. tools/sunset/ (the 34 retired tools) STAYS tracked for the tool index. Contents form, R84.
/docs/sunset/*