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
Risk Level
SAFE
Analyzed
Sep 11, 2026, 02:00 AM
Security Audit — agent-trust-hub — dogfood