copilot-review
Pass
Audited by Gen Agent Trust Hub on Jun 22, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the GitHub CLI (
gh) to retrieve pull request metadata, reviews, and conversation threads. This is consistent with the skill's stated purpose of managing pull request feedback. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data in the form of pull request comments and review summaries.
- Ingestion points: Data enters the agent context via the output of
gh pr viewcommands inSKILL.md. - Boundary markers: No explicit data delimiters are used, but the skill relies on the agent's logic to parse JSON output.
- Capability inventory: The agent can execute shell commands using the
ghtool and perform code modifications based on the feedback. - Sanitization: The skill includes robust defensive instructions requiring the agent to filter feedback by
authorAssociation(e.g., MEMBER, OWNER, COLLABORATOR) and specific bot identities (github-actions[bot],app/github-copilot), effectively ignoring potentially malicious instructions from external contributors.
Audit Metadata