The Agent Skills Directory

[SAFE]: The skill is a well-structured tool that performs local data processing. It uses safe parsing methods (such as yaml.safe_load) and does not utilize risky dynamic execution functions like eval or exec. The skill follows least-privilege principles by working exclusively with local data and providing advisory feedback rather than executing financial transactions.
[INDIRECT_PROMPT_INJECTION]: The skill processes external trade records (JSON/YAML) which constitutes an ingestion surface for indirect prompt injection. Specifically, the review_trade_performance.py script uses the review_id field from the input data to construct output file paths without sanitization. This creates a potential path traversal vulnerability where a crafted input file could cause the script to write reports to unintended local directories.
Ingestion points: The scripts/review_trade_performance.py script reads trade records from files provided via the --input CLI argument.
Boundary markers: No specific delimiters or warnings are used to separate user data from the script's control logic.
Capability inventory: The skill possesses the capability to read and write files to the local filesystem using pathlib and standard file operations.
Sanitization: No sanitization or validation is performed on the review_id or trade_id fields before they are concatenated into output file paths.

trade-performance-coach