afk-architect
Pass
Audited by Gen Agent Trust Hub on Aug 18, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or risky operations were identified. The skill is entirely instructional, focusing on delegating tasks to specific agent roles like 'cartographer', 'builder', or 'pathfinder'.\n- [INDIRECT_PROMPT_INJECTION]: The skill's coordination logic involves processing data from other agents, which is a potential surface for indirect prompt injection. The skill mitigates this through instructions for human-in-the-loop oversight.\n
- Ingestion points: The skill aggregates results from sub-agents ('cartographer', 'builder', 'pathfinder') as described in the orchestration logic within
SKILL.mdandcrew.md.\n - Boundary markers: Explicit instructions require the agent to 'keep approvals with the user' (
SKILL.md) and ensure 'synthesis and approvals' remain in the current thread (agents/openai.yaml).\n - Capability inventory: No file system writes, network operations, or shell command execution capabilities are defined or requested in the skill files.\n
- Sanitization: The framework relies on manual user review and approval of sub-agent outputs before synthesis, preventing automated execution of injected instructions.
Audit Metadata