Explains why YOLO is reasonable inside this specific container (bind-mount
scope, firewall allow-list, non-root user, no host secrets mounted) and
flags the residual risks worth knowing: uncommitted work in /workspace,
exfiltration via GitHub / Anthropic, no CPU/RAM limits. Key advice: commit
before long autonomous runs.
Claude Code running inside the project's dev container can now build the
game, launch a real Godot instance under Xvfb, and drive the automation
harness end-to-end — no Windows dependency.
Dockerfile adds (as root, before USER node):
- X11 / Mesa software GL / audio runtime deps + python3
- .NET SDK 9.0 via upstream dot.net install script -> /usr/local/dotnet
- Godot 4.6.2-stable mono Linux x86_64 -> /opt/godot/godot
- /usr/local/bin/godot-xvfb wrapper: auto-wraps invocations in
xvfb-run -a --server-args="-screen 0 1280x720x24 ..."
harness.py picks GODOT_BIN from env, defaults to /opt/godot/godot on
Linux, and auto-wraps the subprocess in xvfb-run when DISPLAY is unset.
Windows code path unchanged.
init-firewall.sh adds api.nuget.org to the allowlist so dotnet restore
works post-boot. Godot + .NET SDK are fetched at image build time, before
the firewall exists.
New docs:
- autonomous_plan.md: design rationale, alternatives considered
- README.md: launch instructions for Windows terminal / Docker Desktop /
VS Code Dev Containers / WSL2 natif
- CLAUDE.md already documents the harness (done in previous commit)
Validation: docker build succeeds; inside the container, dotnet --version
=9.0.313, godot --version=4.6.2.stable.mono, dotnet test=102/102,
python3 tools/automation/smoke.py passes end-to-end with 14 non-black
1280x720 PNGs. Mission 1 screenshot is visually identical to the Windows
build, and Xvfb determinism is a bonus (det_a.png ≡ det_b.png bytewise).
Launching Godot with --automation=<dir> activates an AutomationHarness
node that polls <dir>/inbox/ for JSON command files, executes them via
a thin facade over existing public surfaces (GameSim, InputMapper,
EventAnimator, ControlBar, PieceStockPanel), and writes results plus
screenshots back to disk. The black-box simulation boundary is not
crossed — every command routes through the same signals/methods a real
player would trigger.
A stdlib-only Python helper (tools/automation/harness.py) wraps the
protocol for test scripts and interactive REPLs. Smoke test passes
end-to-end: load mission, place a piece, step 10 turns, capture 14
1280x720 PNGs, handle rejections, quit cleanly. Existing 102 engine
unit tests still green.
Bundles in-flight work on the campaign/missions system (CampaignDef,
MissionDef, TerrainPatch, TransformerDef, MissionChecker, CampaignLoader,
FlavorBanner, transformer rules), plan files, and matching tests. Baseline
commit so the upcoming automation testing harness lands on a clean tree.
- Mouse wheel up/down zooms camera in/out
- Zoom clamped between 0.3x and 3x to prevent getting lost
- Updated PLAN.md: marked phases 5-6 as done, added phase 7-8
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Dame (Queen):
- Moves 1-2 cells in all 8 directions (orthogonal + diagonal)
- Social status 7 (highest — priority over all other pieces)
- Deep burgundy color, letter "D"
- Rare and powerful, forces strategic placement choices
Levels:
- Level 7 "La Dame Blanche": 10x10, walled arena with central fortress,
1 queen available as a logistics superweapon
- Level 8 "Le Grand Reseau": 12x10, 4 productions + 4 demands,
two vertical wall corridors, 2 queens, full network challenge
GDD updated with Dame section and status hierarchy.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Bugfixes:
- Stop now fully rebuilds piece visuals (fixes destroyed pieces not
reappearing and lingering cargo indicators)
- Piece stock buttons use explicit selected/unselected styling instead
of ambiguous toggle behavior (teal bg + bright border when selected)
- Color dots next to stock buttons match piece colors
Features:
- Clicking a placed piece highlights its start/end cells with piece color
- Back-to-menu button ("← Menu") in top-left returns to level selector
with fade transition
- DetailPanel shows "Pion" name and has styled remove button
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Pion (Pawn): orthogonal range 1, social status 1 (lowest), green color
- All existing levels get surplus stock including pawns for player choice
- Level 4 "Le Carrefour": 8x8, dual cargo diagonal routes, center wall block
- Level 5 "Le Labyrinthe": 8x6, wall corridors, knights essential for shortcuts
- Level 6 "Trois Royaumes": 10x8, 3 productions + 3 demands, full network puzzle
- Production interval concept removed (all produce every turn)
- GDD updated with Pion section and 6 level descriptions
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- New turn order: produce -> transfer -> move -> collision resolution
- Collisions now destroy weaker pieces (status > level > mutual destruction)
instead of halting the simulation. SimPhase.Collision removed.
- Add piece Level property (all level 1 in proto, prepared for future)
- Production fires every turn (interval concept removed), buffer = Amount
(default 1, future 2-4), leftovers overwritten each turn
- Transfer tiebreaker: status > level > clockwise direction (alternating
even/odd turns in y-up coords), replaces distance-to-production
- Demands always accept matching cargo even when already satisfied
- TurnNumber added to all turn events for animation grouping
- Simultaneous animations: produce flash, cargo slide, parallel piece moves
- Camera centering fix + middle-click pan
- GDD updated with new rules + lore section added
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add CargoFilter property to PieceState, auto-assigned at placement
by tracing relay chain back to production. TransferResolver now
enforces cargo-type filtering and uses forward-direction sorting
with cargo-aware distance calculations. Prevents cross-route
contamination on multi-cargo boards.
Level 3 restored to dual-cargo (Wood+Stone) with correct 10R+2K stock.
Two new solvability tests validate filter auto-assignment and chain
propagation. All 60 tests green.
- Fix cargo bouncing in relay chains: piece-to-piece transfers now prefer
receivers farther from production (forward flow) instead of closer (backward)
- Level 2 stock corrected: 4R+1B → 6R+1B to match required solution
- Level 3 simplified to single cargo type (6R+1K) — dual-cargo on 6x6
requires engine support for cargo-type filtering (Phase 2)
- Add PLAN.md with prototype roadmap (phases 2-6)
- 57 tests passing
Black box sim engine (commands in, events out) with 3 piece types
(Rook, Bishop, Knight), cargo transfer system with social status
priority, collision detection, and victory/defeat conditions.
57 tests covering rules, simulation, loading, and solvability.
Godot 4 presentation layer scaffolding.
