senpi-strategy-discover
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The discovery engine fetches a strategy catalog from the vendor's official repository at github.com/Senpi-ai/senpi-skills. This operation uses a well-known service and targets a verified vendor repository to keep strategy information current.- [COMMAND_EXECUTION]: The skill's test suite in tests/test_cli.py utilizes subprocess.run to execute the discovery script, ensuring proper functionality during testing and development.- [DYNAMIC_EXECUTION]: The scripts/discover.py script employs dynamic module loading to import the standard tempfile library, which is used to manage temporary local storage for the catalog.- [DATA_EXFILTRATION]: The skill accesses user budget and holdings data via the vendor's MCP server at mcp.prod.senpi.ai. This communication is required for personalized strategy ranking and remains within the vendor's trusted infrastructure.- [INDIRECT_PROMPT_INJECTION]: The skill processes strategy metadata and caveats, which serves as a potential surface for indirect injection if the source data were compromised. The skill implements a safety-conscious evidence chain:
- Ingestion points: Strategy data is ingested via scripts/discover.py.
- Boundary markers: The skill includes explicit instructions for the agent to report all strategy caveats verbatim, ensuring transparency and user awareness.
- Capability inventory: The skill does not possess capabilities for arbitrary file writes or network communication to non-vendor domains.
- Sanitization: Data is processed through a deterministic Python filtering engine before it reaches the agent context.
Audit Metadata