visual-recap
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local
gitandgh(GitHub CLI) commands to analyze repository diffs and update PR descriptions. These commands are executed via Node.jsexecFileSyncwith array-based arguments, which prevents shell command injection. The tool's access is limited to the local repository and the specific PR being processed. - [INDIRECT_PROMPT_INJECTION]: The skill processes external data, creating a potential injection surface.
- Ingestion points: Fetches PR body text using
gh pr viewinscripts/upsert-recap-block.mjs. - Boundary markers: Uses HTML comments
<!-- system-recap:start -->and<!-- system-recap:end -->to delimit its output. - Capability inventory:
gh pr editfor PR updates,git difffor repository inspection, and Node.js script execution. - Sanitization: The
mergeRecapBodyfunction performs string replacement only within the defined marker boundaries, protecting the surrounding PR content.
Audit Metadata