analysis-technical
Pass
Audited by Gen Agent Trust Hub on Jun 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The script "scripts/ta_client.py" contains logic to execute a local companion script ("yfinance_client.py") using "subprocess.run" to fetch market data. The implementation uses a list-based argument structure without a shell, which is a secure practice that prevents command injection.
- [EXTERNAL_DOWNLOADS]: The skill utilizes standard, well-known Python libraries including "pandas" and "numpy". These are declared in the script headers for environment setup via "uv".
- [PROMPT_INJECTION]: The skill processes external market data in JSON format. While this constitutes an attack surface for indirect prompt injection, the skill mitigates risks by strictly parsing the JSON and casting price/volume data to floating-point numbers before performing any calculations.
- Ingestion points: Ingests OHLCV JSON data via "--input" or stdin in "scripts/ta_compute.py" and "scripts/ta_client.py".
- Boundary markers: Data is parsed programmatically; no natural language boundary markers are used as the data does not interact with the LLM prompt directly during calculation.
- Capability inventory: "subprocess.run" is present in "ta_client.py" for local script execution but is not triggered by the data ingestion path.
- Sanitization: Price and volume fields are explicitly converted to float types, preventing non-numeric data from influencing the logic.
Audit Metadata