The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script (pmf_score.py) to calculate performance gaps and status for product metrics. This script is contained within the skill's own directory, uses only standard library modules, and does not perform network operations or persistent file modifications.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from user pastes and uploaded files. It contains robust defensive instructions to treat all fetched or uploaded text as data rather than instructions, explicitly directing the agent to ignore and report any command-like text found within processed data.
[DATA_EXPOSURE]: While the skill reads sensitive business metrics, it does not include any network-capable code or exfiltration patterns. Data processing is handled locally via the included script or within the agent's context, and results are returned directly to the user.
[DYNAMIC_EXECUTION]: The skill uses Python to process user data at runtime. This is implemented safely by converting user input into a structured JSON object which is then passed to a static script, minimizing the risk of arbitrary code execution from malicious inputs.

pmf-check