1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
/target
.cargo/
!.cargo/config.toml
npm/ilo.wasm
.env
__pycache__/
/ilo
*.png
.playwright-mcp/
.worktrees
.claude/worktrees/
# research/ is untracked by default (scratch space + private explorations).
# Track the closed-loop benchmark explicitly - it is a publishable artefact.
research/*
!research/closed-loop-bench/
!research/BENCHMARK-METHODOLOGY.md
!research/finetune-comparison.md
# PNGs in the benchmark charts/ are dataset artefacts and must be tracked.
!research/closed-loop-bench/charts/*.png
# Per-cell raw results and tool caches are reproducible from the driver -
# don't bloat the repo with them.
research/closed-loop-bench/results/raw/
research/closed-loop-bench/.zero/
research/closed-loop-bench/**/__pycache__/
site/
.DS_Store
.astro/
node_modules/
rust_out
# Research / scratch docs (subagents leave reports here; not tracked)
docs/
# Tracked documentation files are explicitly exempted below:
!docs/deploying.md
!docs/streaming.md
# E2E test artefacts (subagent-produced logs; not tracked)
tests/e2e/
# CLAUDE.md is per-developer Claude Code context; AGENTS.md is the cross-tool tracked equivalent
CLAUDE.md
# Internal strategy and exploration (kept local, not public)
zero-gap-specs/
!research/sigil-ab-results.md
!research/sigil-ab/
!research/sigil-ab/**
.claude/commands/
.env.example
.zero/