The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data sources, including historical PE CSV files (history_csv) and analyst consensus data. Maliciously crafted data in these sources could potentially influence the agent's behavior during calculation processing.
Ingestion points: Processes external files via raw_ref and history_csv parameters, as well as analyst forecasts (eps_forecast).
Boundary markers: The skill requires an input evidence id, but does not specify explicit delimiters or "ignore instructions" wrappers for the data content itself.
Capability inventory: The agent is authorized to execute a local Python script calc/cli.py to perform calculations.
Sanitization: No specific sanitization or validation logic is described for the external data before it is passed to the calculation tool.
[COMMAND_EXECUTION]: The skill's primary function involves invoking a local CLI tool (calc/cli.py) with arguments derived from user queries and external data. While this is the intended purpose, it creates a surface for command argument injection if input data is not properly handled by the underlying tool.

valuation