momentic-maintain
Pass
Audited by Gen Agent Trust Hub on Nov 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external test artifacts such as logs, DOM snapshots, and error traces to diagnose failures. This represents an attack surface where malicious or deceptive test output could attempt to influence the agent's reasoning during the classification or repair process.
- Ingestion points: Run metadata, screenshots, and logs fetched from the Momentic API (
api.momentic.ai) or local ZIP artifacts. - Boundary markers: The instructions include a reasoning framework requiring specific evidence (screenshots, traces) to support every causal claim and diagnostic verdict.
- Capability inventory: The skill possesses the ability to perform network requests to vendor APIs, execute the
momenticCLI, and modify test files on the local filesystem. - Sanitization: The instructions emphasize human review and approval for repairs and explicitly warn against committing sensitive data like signed URLs or credentials.
Audit Metadata