lov-skill-add-case
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The script
scripts/add_case_with_session.pyutilizessubprocess.runto execute an external helper script (share_session.py). The command is constructed as a list using the current Python executable and a resolved script path, following best practices for avoiding shell-related injection vulnerabilities. - [INDIRECT_PROMPT_INJECTION]: The skill processes 'accepted results' (data generated by other AI interactions) to format them into website-compatible JSON. This constitutes a data ingestion surface.
- Ingestion points: Data enters the skill via JSON objects passed to the
--caseand--submissionarguments inscripts/add_case.py,scripts/submit_case.py, andscripts/add_case_with_session.py. - Boundary markers: The workflow strictly requires the input data to include an explicit user-confirmation flag (
"acceptance": "user-confirmed") and includes a 'preview' step before publication. - Capability inventory: The skill can perform network operations (limited to
lovstudio.ai), write local files (updating the case registry and credential cache), and execute local Python scripts via subprocesses. - Sanitization: The skill implements comprehensive regex-based scanners (
SECRET_PATTERNSandSECRET) inscripts/add_case.pyandscripts/submit_case.pyto identify and block the inclusion of sensitive data such as API keys, private filesystem paths, and authentication tokens in the public case bundle.
Audit Metadata