The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from user-provided JSON files which are processed by scripts to generate market research reports.\n
Ingestion points: Data enters the agent's context through the --input flag in tam_sam_som_builder.py, demand_signal_triangulator.py, and survey_instrument_auditor.py as described in the SKILL.md workflows.\n
Boundary markers: The skill relies on structured JSON formatting for input data, though it lacks explicit "ignore embedded instructions" markers when the agent interprets the resulting script output.\n
Capability inventory: The scripts are restricted to mathematical computation and regex-based auditing. They contain no subprocess calls, network operations, or dynamic code execution (eval/exec).\n
Sanitization: Inputs are validated for correct JSON structure and expected value ranges (e.g., probability ratios between 0 and 1) within the Python script logic.\n- [SAFE]: No malicious patterns, unauthorized network access, or sensitive data exposure were detected in the skill's instructions or scripts. The Python tools use only standard library modules and perform deterministic operations.

market-research