siyu-pyq
Warn
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDYNAMIC_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions direct the agent to execute a local Python script (pyq_lint.py) via a shell command to perform compliance checks on generated content.
- [DYNAMIC_EXECUTION]: The scripts/pyq_lint.py script manipulates sys.path using relative paths derived from the script location (parents[5] and parents[3]) to dynamically load and import the siyu_team package.
- [DATA_EXFILTRATION]: The skill is configured to read from files located in ~/.siyu/clients/ to restore client profiles. Accessing files in the user home directory is a sensitive operation.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data from user-provided materials and client files.
- Ingestion points: Files in ~/.siyu/clients/ and user-provided 'real materials' in assets/真实语料-待填.md.
- Boundary markers: None present in instructions.
- Capability inventory: Shell command execution (pyq_lint.py) and local file read.
- Sanitization: Compliance scanning for prohibited keywords is present in pyq_lint.py, but it does not address prompt injection specifically.
Audit Metadata