boss-advanced

Pass

Audited by Gen Agent Trust Hub on Jul 25, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill defines advanced orchestration logic, delegation templates, and conflict resolution strategies for managing multiple AI agents.
  • [INDIRECT_PROMPT_INJECTION]: The orchestration framework relies on "Guardian" agents processing outputs from other "Implementer" agents, creating a potential surface for indirect prompt injection if an implementer's output contains malicious instructions.
  • Ingestion points: Guardian agents ingest data from implementers via SendMessage or shared task lists as described in SKILL.md.
  • Boundary markers: The provided instructions do not specify explicit delimiters or isolation markers for inter-agent communication.
  • Capability inventory: The framework utilizes agent spawning and git commands (git worktree) for task execution.
  • Sanitization: No explicit sanitization or verification steps for inter-agent messages are detailed in the delegation logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 25, 2026, 08:09 PM
Security Audit — agent-trust-hub — boss-advanced