Skip to content

meshmapper CLI

meshmapper fuses saved artifacts into one graph, runs the M4 rule pack, and writes unvalidated path hypotheses for operational targeting.

cd agentic-redteam/meshmapper
meshmapper \
  --transcript ../seam/transcript.json \
  --aipostex endpoints.json \
  --agent-card billing-card.json \
  --mcp-tools-list tools.json \
  --config mesh.yaml \
  --graph graph.json \
  --out paths.json \
  --schema ../schema

At least one input source is required.

Inputs

All source flags may be repeated:

  • --transcript: Seam transcript.json.
  • --aipostex: endpoint inventory JSON.
  • --agent-card: standalone A2A Agent Card JSON.
  • --mcp-tools-list: MCP tools/list JSON-RPC result or tool array.
  • --config: YAML or JSON orchestration config.

Outputs

  • --graph: deterministic serialized graph, defaulting to graph.json.
  • --out: path-hypothesis output, defaulting to paths.json.
  • --schema: schema directory or direct path-hypothesis.schema.json path, defaulting to ../schema.

The command validates emitted hypotheses before writing them. A successful run prints the graph path and the number of emitted hypotheses.

Install

cd agentic-redteam/meshmapper
python3 -m pip install -e .
python3 -m pytest -q

See ingestion for accepted source shapes and hypotheses for output semantics.

Robustness Scenarios

M5 adds deterministic topology checks:

cd agentic-redteam/meshmapper
meshmapper robustness run \
  --scenario scenarios/hierarchical_planner_executor.yaml \
  --out bundles/hierarchical \
  --schema ../schema

The bundle contains scenario.yaml, graph.json, paths.json, and results.json. Built-in scenarios cover hierarchical planner-executor, peer-swarm trust spoofing, blackboard/memory propagation, and a safe authenticated negative topology.