pr-stale-sweep
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- Command Execution: The skill interacts with the repository environment through the GitHub command-line interface to list, comment on, and close pull requests. These actions are performed sequentially and require human review.
- Dynamic Execution: A local Python module is executed during a 'pre-flight' phase to verify that the environment is correctly configured and that skill hashes match expected values. This is used for framework integrity and configuration verification.
- Indirect Prompt Injection Surface: By reading pull request titles, bodies, and comments, the skill ingests potentially untrusted external data. It explicitly identifies this risk and mandates that such content be treated strictly as data rather than instructions, mitigating potential prompt injection attacks.
- Security-Conscious Workflow: The skill implements specific safeguards, such as skipping pull requests that mention potential vulnerabilities (e.g., CVE references) to prevent public disclosure of sensitive information via automated comments.
- Command Injection Mitigation: To prevent shell injection when posting comments, the skill utilizes temporary files to pass comment content to the command-line tool instead of direct string interpolation, ensuring that external data does not influence the shell execution environment.
Audit Metadata