agent-protocol

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [NO_CODE]: The skill consists entirely of Markdown documentation and instructional templates for agent interaction. No scripts, binaries, or automated installation procedures are present.
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a framework for processing data generated by other agents, creating a theoretical surface for inter-agent instructions.
  • Ingestion points: The protocol uses [INVOKE] and [RESPONSE] tags in SKILL.md to define how agents receive and process queries from other roles.
  • Boundary markers: The protocol mandates clear delimiters, including closing tags like [/RESPONSE] and structured chain tracking [CHAIN: ...] to maintain context isolation.
  • Capability inventory: There are no capabilities for file writing, network requests, or command execution defined within the skill.
  • Sanitization: The skill includes a robust 'Internal Quality Loop' requiring multi-stage verification (Self-Verification, Peer-Verification, and Critic Pre-Screen) to identify and flag assumptions or contradictions before output is generated.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 12:31 PM
Security Audit — agent-trust-hub — agent-protocol