crew
Pass
Audited by Gen Agent Trust Hub on Jul 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill dynamically executes verification commands retrieved from the
RIGGING.mdconfiguration file.\n- [PROMPT_INJECTION]: The skill processes external data including failure logs, feature specifications, and repository-level configuration files, creating a surface for indirect prompt injection.\n - Ingestion points: The agent reads
RIGGING.md,AGENTS.md, feature files, and failure evidence (runner output) into the active context.\n - Boundary markers: The instructions do not define specific delimiters or warnings to ignore instructions embedded within the content of these external files.\n
- Capability inventory: The agent has the capability to modify production source code and execute shell commands defined in configuration files.\n
- Sanitization: The skill provides no mechanisms for the validation or sanitization of content found in the ingested logs or specifications.\n- [COMMAND_EXECUTION]: The instructions explicitly direct the agent to avoid using defensive error handling mechanisms (such as try/catch blocks or error-wrapping types) in production code unless a failure scenario specifically mandates them. This instruction results in the production of code that may be more vulnerable to denial-of-service or information disclosure via unhandled exceptions.
Audit Metadata