hyperplan
Pass
Audited by Gen Agent Trust Hub on Aug 1, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill implements a multi-agent orchestration pattern that ingests untrusted user input (planning requests) and propagates it to multiple sub-agents (skeptic, validator, researcher, architect, and creative). This architecture creates a surface for indirect prompt injection where a malicious user request could attempt to subvert the adversarial roles or the synthesis logic.
- Ingestion points: Phase 2 of the workflow captures the user's planning request and interpolates it into prompts sent to five different team members.
- Boundary markers: The skill utilizes
<user-request>and</user-request>XML-style delimiters to wrap the untrusted user content. - Capability inventory: The orchestrator employs the
team_send_messagetool for inter-agent communication and thetasktool to dispatch a final formalization task to aplansubagent. - Sanitization: There is no evidence of explicit input validation, escaping, or filtering of the user's input before it is interpolated into the system prompts of the team members.
Audit Metadata