interface-review
Pass
Audited by Gen Agent Trust Hub on Nov 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted content from Pull Request metadata and repository history, which could contain malicious instructions designed to influence the agent's behavior.
- Ingestion points: Pull request titles and bodies fetched via
gh pr view, commit messages, and source code content accessed throughgit showandgit diff(SKILL.md, Section 6 and scope-resolution.md). - Boundary markers: The instructions do not specify the use of delimiters or specific "ignore" blocks to isolate untrusted PR metadata from the agent's execution context.
- Capability inventory: The skill has access to shell commands (
git,gh,grep), performs network fetches from GitHub, and can create temporary directories (git worktree add). - Sanitization: There is no evidence of sanitization or escaping of external text before it is processed by the agent.
- [COMMAND_EXECUTION]: The skill dynamically generates and executes a variety of shell commands to inspect the repository state and resolve change scopes.
- Evidence: Use of
git diff,git blame,git show,git grep,git worktree, andghCLI commands acrossSKILL.mdandscope-resolution.mdto analyze the codebase. - [EXTERNAL_DOWNLOADS]: The skill performs network operations to fetch code and metadata from GitHub. These operations target a well-known and trusted service.
- Evidence:
git fetch origin "pull/<n>/head:..."andgh pr vieware used to retrieve remote content for analysis (scope-resolution.md).
Audit Metadata