dogfood
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a shell command to resolve the comparison base using a script located at
${CLAUDE_PLUGIN_ROOT:-.}/scripts/resolve-pr-base.sh. This script is a local resource intended for development workflows. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content, creating a vulnerability surface where malicious instructions could be embedded in data or bug reports.
- Ingestion points: Pull request content, "reporter's exact steps" from bug reports, and "live-scale data" used for testing (SKILL.md).
- Boundary markers: Absent. The instructions do not specify using delimiters or cautioning the agent about instructions within the processed data.
- Capability inventory: The skill allows the agent to execute CLI commands, make network requests (API/worker entrypoint), and modify code via the Repair/TDD loop.
- Sanitization: Absent. The skill instructs the agent to follow reporter steps exactly, which is a common vector for instruction injection.
Audit Metadata