loop-address-pr-feedback
Pass
Audited by Gen Agent Trust Hub on Feb 23, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it fetches review feedback from external sources and processes it using a code-generation agent.
- Ingestion points: The skill fetches review comments and issue comments via the GitHub API (
/issues/{num}/commentsand/pulls/{num}/comments) in theGatherphase. - Boundary markers: The
Fixphase task prompt (Fix the issue found by code review: ...) does not include delimiters or instructions for the sub-agent to ignore potential instructions embedded within the feedback. - Capability inventory: The agent has the capability to write code, commit changes, and push to remote repositories, which provides a direct path for executing injected instructions.
- Sanitization: No sanitization or validation logic is present to filter malicious content from the review comments before processing.
- [COMMAND_EXECUTION]: The skill uses several command-line tools to interact with the repository and the GitHub API.
- Evidence: Executes
gt(Graphite CLI) for stack management and log analysis. - Evidence: Uses
gh(GitHub CLI) for PR list and API interactions. - Evidence: Performs standard
gitoperations (add,commit,push) to manage the repository state.
Audit Metadata