gh-fix-ci
Pass
Audited by Gen Agent Trust Hub on Mar 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The bundled script
scripts/inspect_pr_checks.pyuses the Pythonsubprocessmodule to executegitandgh(GitHub CLI) commands. These commands are used to resolve Git repository roots, view pull request metadata, list CI check statuses, and fetch job logs. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from external CI logs and has the capability to modify the repository.
- Ingestion points: Build logs are retrieved from GitHub Actions runs in
scripts/inspect_pr_checks.pyusinggh run view --logand the GitHub API. - Boundary markers: Logs are provided to the agent as text snippets or JSON without explicit boundary markers or instructions to disregard embedded commands.
- Capability inventory: The workflow defined in
SKILL.mdexplicitly allows the agent to "commit and push changes" to the repository based on its analysis of the logs. - Sanitization: There is no evidence of sanitization or filtering of the log content to remove potential injection strings before the data is processed by the LLM.
Audit Metadata