gh-cli
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides comprehensive patterns for driving the official GitHub CLI (
gh). It includes examples for managing pull requests, issues, releases, gists, Actions, and repository metadata. - [INDIRECT_PROMPT_INJECTION]: The skill explicitly addresses and provides mitigations for potential injection attacks originating from untrusted repository content.
- Ingestion points: Data fetched from
gh gist view,gh apiresponses,gh pr diff, release artifacts, and codespace logs (SKILL.md). - Boundary markers: The instructions mandate avoiding terminal-rendered output and instead recommend using
--jsonflags or redirecting raw content to files to prevent ANSI escape sequence injection. - Capability inventory: The skill utilizes the
ghtoolset for extensive repository interaction, including API calls, file reads, and CI/CD management. - Sanitization: Guidelines instruct the agent to use
--jsonfields for structured parsing and to redact token fragments fromgh auth statusoutput. - [EXTERNAL_DOWNLOADS]: The skill documents the process for installing GitHub CLI extensions and agent skills, which involves downloading code from external repositories.
- Evidence: Mentions
gh extension install github/gh-awandgh skill install owner/repo --agent opencode. These are standard functionalities of the GitHub CLI ecosystem. - [SAFE]: The content acts as a security-aware reference, highlighting known GitHub Security Advisories (e.g., GHSA-cg6r-mpgc-h9mm, GHSA-mm27-mwq9-fr5g) and establishing best practices to ensure the agent does not inadvertently compromise the host environment or expose credentials.
Audit Metadata