The Agent Skills Directory

[COMMAND_EXECUTION]: The skill interactively executes gh and git commands using subprocess.run. The implementation follows security best practices by passing arguments as a list rather than a single shell string, effectively preventing shell injection vulnerabilities.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and processes GitHub Actions logs, which may contain attacker-influenced content.
Ingestion points: Untrusted data enters via gh run view --log and gh api calls in scripts/inspect_pr_checks.py.
Boundary markers: There are no explicit delimiters or instructions to ignore embedded commands in the processed logs.
Capability inventory: The skill includes the ability to execute CLI tools and modify the local filesystem during the fix implementation phase.
Sanitization: The skill does not sanitize the logs before they are presented to the AI agent.
Mitigation: The workflow mandates that any fix must be reviewed and approved by the user before implementation, which serves as a critical safety barrier.

gh-fix-ci