agentic-loop-github
Pass
Audited by Gen Agent Trust Hub on Jul 14, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill defines an 'observe-decide-execute-verify' workflow for GitHub tasks. It utilizes legitimate tools such as the GitHub CLI (
gh) andgitfor repository automation and state management. - [PROMPT_INJECTION]: The skill functions by ingesting external data, specifically GitHub CI logs and PR check outputs, to drive its decision-making loop. This creates an interface for potential indirect prompt injection if the ingested data contains instructions designed to influence the agent. The skill includes design patterns like stagnation detection and objective verification to limit the impact of such data.
- Ingestion points:
gh pr checks, CI logs, and repository diffs are read into the agent context. - Boundary markers: The skill does not define specific delimiters for separating external logs from the main instruction context.
- Capability inventory: The skill utilizes file system writes,
git push,git checkout, andgit resetcommands. - Sanitization: There are no specific instructions for filtering or sanitizing the content of the ingested logs.
Audit Metadata