The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local script located at tools/open_loops.py to retrieve task statuses. This execution is part of the skill's primary function and does not involve remote downloads or arbitrary command injection from external sources.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from local project files, creating a potential surface for processing untrusted content if those files are compromised. However, the risk is managed as the sources are internal project documents, and the skill includes instructions for evidence-based assessment.
Ingestion points: context/priorities.md, decisions/log.md, and the projects/ directory.
Boundary markers: Not present.
Capability inventory: Local subprocess execution of tools/open_loops.py and file-write access to decisions/log.md.
Sanitization: Instructions mandate "evidence over assertion" and specify that no confidential figures should be persisted, directing the agent to reference live sources instead.

weekly