maintain-verification-skill
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources (repository files) and uses it to drive execution logic.
- Ingestion points: The agent reads command strings and expectations from the project's verify skill and feature map files.
- Boundary markers: No markers are defined to isolate untrusted command strings from the execution context.
- Capability inventory: The agent is granted permission to execute shell commands found in these files and perform file-write operations via a ledger package.
- Sanitization: No validation or filtering of command strings is performed before execution.
- [COMMAND_EXECUTION]: The workflow requires the agent to execute shell commands discovered in the project's metadata ('execute each surviving command') to synchronize state with the current repository reality.
Audit Metadata