fix-renovate-pr
Pass
Audited by Gen Agent Trust Hub on Jun 28, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: Indirect prompt injection surface detected. The skill processes untrusted data which could contain instructions aimed at manipulating agent behavior.
- Ingestion points: PR body text, issue comments, and CI job logs retrieved via GitHub MCP tools.
- Boundary markers: None present in the instructions to delimit external data from agent instructions.
- Capability inventory: File system modification,
npmcommand execution, and execution of arbitrary commands found in CI logs. - Sanitization: No explicit sanitization or validation of the PR content or logs is performed before processing.
- [COMMAND_EXECUTION]: The skill directs the agent to execute commands derived from CI logs to reproduce failures and verify fixes. It also specifically uses
npm ciandnpm installfor dependency management. These actions are restricted to the local development environment and are necessary for the skill's primary function.
Audit Metadata