orchestration

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [Indirect Prompt Injection] (LOW): The skill's delegation protocol involves passing raw external data directly into the prompts of specialized sub-agents.
  • Ingestion points: Untrusted data enters the context via the 'ERROR' (exact error messages) and 'REQUIREMENTS' fields in the 'Required context template' within SKILL.md.
  • Boundary markers: The template uses simple text headers (e.g., ERROR:, TASK:) which are insufficient to prevent an adversary from escaping the field and injecting instructions.
  • Capability inventory: While the orchestrator itself is restricted from writing files, the roles it manages (Implementer, Test Writer, File Updater) possess Write and Edit capabilities on the filesystem.
  • Sanitization: The skill does not prescribe any escaping, filtering, or validation for the external content interpolated into the delegation templates.
  • [External Downloads] (LOW): The 'Dependencies' section recommends using npx skills add to install skills from jwilger/agent-skills, which is not an organization or repository on the trusted sources list.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 09:41 PM