The Agent Skills Directory

[SAFE]: The customer_interview_analyzer.py script implements a robust safe_user_path function that prevents path traversal attacks by ensuring all file access is restricted to the current workspace.
[SAFE]: The rice_prioritizer.py script includes a remediation for CSV injection (DDE) by prepending a single quote to cells starting with dangerous characters (=, +, -, @) when exporting data to CSV.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from interview transcripts and feature list CSVs which could theoretically contain instructions intended to influence the agent's analysis or output.
Ingestion points: Reads files provided via command-line arguments in scripts/customer_interview_analyzer.py and scripts/rice_prioritizer.py.
Boundary markers: None; the content is processed as raw text or CSV rows.
Capability inventory: Limited to local file read and write; no network access or arbitrary code execution capabilities are present in the scripts.
Sanitization: Includes specific protection for CSV injection in output files, and the SKILL.md file explicitly warns users to review all extracted quotes and themes against original transcripts to verify accuracy.

product-manager-toolkit