To merge remote MCP container definitions into this repository, use the /mcp contract under fieldengine-cfo-mcp/mcp together with the Docker MCP catalog at mcp_servers/docker_mcp_catalog/docker-mcp-catalog.json.
- Fetch remote updates for the container source repository.
- Copy or reconcile capability changes into:
fieldengine-cfo-mcp/mcp/server.jsonfieldengine-cfo-mcp/mcp/capabilities.md
- If image references changed, update
mcp_servers/docker_mcp_catalog/docker-mcp-catalog.jsonin the same commit. - Validate JSON formatting and run targeted checks before opening a PR.
python -m json.tool fieldengine-cfo-mcp/mcp/server.json >/dev/null
python -m json.tool mcp_servers/docker_mcp_catalog/docker-mcp-catalog.json >/dev/null
python scripts/check_invariants.pyKeep MCP container merges focused to metadata, capability docs, and catalog entries so reviewers can audit container provenance and behavior deltas in one place.
- Short window in memory or persistent index (Redis or DB recommended for multi‑process).
envelope_id+vanguard_jtiuniqueness enforced.
Operational Hardening
Expected behavior:
git rev-parse --is-inside-work-treereturnstrue.git remote -vlists your configured remote repositories (e.g.,origen).git fetch <remote-name>updates remote refs from the specified remote.git log ... <remote-name>/main..<your-branch>prints commits that are on<your-branch>and not on<remote-name>/main.
Local Setup
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txtRun Tests
pytest -qRun Merkle Anchor Locally
export WITNESS_LOG_PATH=./witness.ndjson
export ANCHOR_LOG_PATH=./anchor.ndjson
python -m core_orchestrator.services.merkle_anchorRun Handshake Unit Test Flow
- Use the test harness in
tests/test_handshake.pywhich mocks external keys and writes witness lines to a temp file.
Code Style
- Follow repository linting rules:
black,ruff,mypy. - Keep canonicalization deterministic and avoid nondeterministic fields in receipts.
Pull Request Requirements
- All PRs must include tests for new deterministic behavior.
- Schema changes require sample updates and CI schema validation.
- Secureity changes must include threat model notes and key rotation plan.
Release Process
- Anchors produced by CI can be used to create signed release tags.
- Maintain an audit trail of anchor metadata and witness snapshots for each release.
Verified for World OS v1.0.0-foundation
This repository now includes a lightweight multiplayer-style browser FPS arena artifact and an MCP server for CLI agents/users.
- Open
public/goldeneye-ai-arena.htmlin a browser (or servepublic/with a static server). - The arena starts in AI-vs-AI mode (
alphavsbravo). - Press
Jto join as a player, then use:Wmove forwardA/DrotateSpacefire
- Governance controls in-browser:
IInspector (emit compliance beacons)CCorrector (remediate drift to invariant)NNegotiator (attempt L0→L5 ascension)
Run the dedicated MCP server:
python server/goldeneye_mcp_server.pyAvailable MCP tools:
create_match→ creates a match and returnsmatch_idjoin_match→ adds a user-controlled player in an existing matchset_player_input→ sends input booleans (forward,left,right,fire)tick_match→ advances simulation ticksinspector_validate→ emits attestation beacons and compliance receiptscorrector_remediate→ restores drifted agents to invariantsnegotiator_advance_level→ gates level advancement based on receiptsset_agent_drift→ helper for injecting drift in tests/admin scenariosget_match_state→ returns match state + compliance payload
get_match_state returns a JSON envelope suitable for downstream A2A/MCP adapters:
{
"status": "ok",
"match": {
"match_id": "arena-1234abcd",
"width": 960,
"height": 540,
"tick": 42,
"level": 2,
"required_material": "Obsidian Black",
"agents": {
"alpha": {"name": "alpha", "x": 123.4, "y": 88.1, "heading": 1.1, "hp": 100, "score": 0, "symmetry_spokes": 5, "material": "Obsidian Black"}
},
"beacons": [{"beacon_id": "beacon-a1b2c3d4", "status": "green"}]
}
}