The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script to perform filesystem checkpointing and state management.
Evidence: The instruction in SKILL.md uses python3 <skill-dir>/scripts/p2p_filesystem.py to manage checkpoints in the .p2p/ directory.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content from tracker issues and project files, creating an injection surface.
Ingestion points: The skill reads from external tracker issues, issue comments, and repo paths like specs/ or work/.
Boundary markers: The instructions mandate the use of a specific "Acceptance matrix" schema to structure and isolate the ingested data.
Capability inventory: The skill can write files to the .p2p/ directory and execute local management scripts.
Sanitization: The process relies on an agent-led audit for material findings and ambiguity rather than programmatic sanitization.
[DYNAMIC_EXECUTION]: The skill performs dynamic execution of a Python script at runtime for specialized filesystem operations.
Evidence: python3 <skill-dir>/scripts/p2p_filesystem.py is executed as part of the contract handoff and checkpointing process.

plan-acceptance