llp-maintain
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from the local working tree, git diffs, and the documentation corpus to generate maintenance proposals. This ingestion of external content creates a potential surface where malicious instructions embedded in project files could influence the agent's reasoning.
- Ingestion points: Working tree files, git diff ranges, and the LLP documentation corpus.
- Boundary markers: The skill does not specify the use of delimiters or 'ignore' instructions for the content it parses from the corpus.
- Capability inventory: The skill identifies drift and proposes repairs or status changes; it does not perform automated file writes or data exfiltration.
- Sanitization: There is no evidence of input validation or content filtering for the data read from the local environment.
- [DATA_EXFILTRATION]: The skill performs network operations to check for software updates by comparing local source version pins against upstream repositories.
- Evidence: The skill checks if a
source:pin lags behind upstream releases when network access is available.
Audit Metadata