The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or suspicious behaviors were detected during the analysis of the skill's instructions, scripts, or documentation.
[INDIRECT_PROMPT_INJECTION]: The skill provides tools that ingest untrusted data from external sources (interview transcripts and feature CSV files).
Ingestion points: scripts/customer_interview_analyzer.py reads user-provided transcript files; scripts/rice_prioritizer.py reads feature data from CSV files.
Boundary markers: No specific LLM prompt boundary markers are used, as the processing is handled by local Python logic.
Capability inventory: The scripts are limited to file reading, CSV/JSON parsing, and regular expression matching. No network capabilities, sensitive file system access, or subprocess execution were identified.
Sanitization: The scripts use standard Python libraries (csv, json, re) for parsing, which provides inherent structure validation for the inputs.

product-manager-toolkit