The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to run local diagnostic and simulation scripts provided within the user's own project environment to verify system state.
Evidence: repo-runtime.md and eval-cases.md mention the execution of python3 ~/oneos-prod/scripts/repo-sim-multi-ms.py and scripts/prod-ro-probe.sh for environment simulation.
[EXTERNAL_DOWNLOADS]: The skill includes installation instructions that fetch the skill content from the author's public repository.
Evidence: INSTALL.md provides an npx skills add command targeting the repository 15810879921-coder/oneos-pm-skills.
[PROMPT_INJECTION]: The skill defines a robust handling process for its input 'handoff packages' to mitigate indirect injection risks through structural enforcement.
Ingestion points: The agent processes external instructions via a specific markdown schema defined in handoff-from-pm.md.
Boundary markers: The skill employs 'Hard Gates,' which involve rejecting any input that does not conform to the mandatory schema or contains ambiguous instructions.
Capability inventory: The agent has permissions to modify local files in prototype and production directories and execute project-specific scripts.
Sanitization: The instructions strictly prohibit 'brain-filling' or autonomous decision-making, requiring the agent to stop and ask the user for clarification if the input package is incomplete.

mingjingzhishui