trace
Pass
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill directs the agent to execute the application's existing codebase and potentially start local development servers to capture runtime behavior.
- [DYNAMIC_EXECUTION]: The agent is instructed to generate and run temporary scripts that import and invoke local repository modules to drive specific scenarios.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from network calls and database logs, which could contain embedded instructions intended to influence the agent's behavior during the trace write-up.
- Ingestion points: Captured HTTP request/response bodies, database query results, and application console output.
- Boundary markers: No explicit delimiter or 'ignore' instructions are provided in the trace template for the captured external data.
- Capability inventory: Filesystem write access to the repository, shell command execution, and script evaluation.
- Sanitization: Instructions are provided to redact keys and tokens before recording any captured data to disk.
- [DATA_EXFILTRATION]: The skill captures and logs full network payloads and database queries to files within the repository, creating a risk of accidental PII or sensitive application data exposure if redaction is incomplete.
Audit Metadata