swe-orchestrator

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and acts upon data from external VCS issues.
  • Ingestion points: The SWE Architect persona reads issue content using the vcs-issue-management tool (SKILL.md).
  • Boundary markers: Absent; there are no instructions to use specific delimiters or to disregard instructions found within the issue description.
  • Capability inventory: The skill can create and switch branches (vcs-branch), write and commit code (vcs-commit), create pull requests (vcs-change-request), and execute tests or linters.
  • Sanitization: Absent; the agent is directed to gather goals and criteria directly from the issue body without verification.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 02:27 AM