calibrate-first
Pass
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill is designed to run a local Python script named
assay_lite.pyto perform financial calculations. To mitigate risks associated with command execution, the instructions explicitly require that the script be manually reviewed, stored at an absolute path, and verified against a specific cryptographic commit hash (78104a4356541affe2a469dfdf69cb1e418eb0c5). - [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied CSV files (
prices.csv,signal.csv). This ingestion of untrusted data constitutes an attack surface where malicious content in the data could attempt to influence the agent's logic. - Ingestion points: CSV files loaded via the
--pricesand--signalarguments inSKILL.md. - Boundary markers: None explicitly defined for the contents of the CSV files.
- Capability inventory: The skill executes shell commands via Python and writes to a local
.assay-lite/directory as described inSKILL.md. - Sanitization: The instructions do not specify input validation or escaping for the data within the CSV files, relying instead on the integrity of the offline Python tool.
Audit Metadata