slop-fix-issue
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from GitHub issue bodies, discussion threads, and change histories which could contain malicious instructions aimed at overriding agent behavior.
- Ingestion points: GitHub issue state, body, discussion threads, and repository code history (SKILL.md).
- Boundary markers: The skill does not define explicit boundary markers or delimiters to isolate untrusted issue content from the system instructions.
- Capability inventory: The skill utilizes the
ghCLI for issue management and PR creation, and is authorized to execute local validation and regression checks (SKILL.md). - Sanitization: There are no specific instructions for sanitizing or escaping the content retrieved from GitHub before it is processed or used in decision-making.
- [COMMAND_EXECUTION]: The workflow requires the agent to run repository-specific validation and regression checks (SKILL.md). This involves executing arbitrary commands or scripts found within the target repository, which is a powerful capability that could be abused if the repository contains malicious build or test scripts.
Audit Metadata