string-diagram-rewriting-protocol
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides conceptual and mathematical definitions for a string diagram rewriting protocol. It does not include active scripts, executables, or network-enabled tools. The implementation sketches in Python, Julia, and Zig are illustrative and non-functional within the skill itself.
- [INDIRECT_PROMPT_INJECTION]: The protocol is designed to process 'Input Streams' (Section 7.1) which could originate from untrusted external sources. While the protocol itself is just a specification, it creates a surface where malformed data could attempt to influence the agent's decision-making if processed without validation at the implementation layer.
- Ingestion points: Untrusted data enters the rewriting pipeline through the defined 'Input Stream' (Section 7.1).
- Boundary markers: The protocol does not currently define specific delimiters or boundary markers to distinguish between data content and transformation instructions.
- Capability inventory: The protocol connects to Layer 2 and Layer 5 skills, which are responsible for structural rewriting and commitment to actions/decisions.
- Sanitization: No specific sanitization or escaping mechanisms for external content are defined in the protocol specification.
Audit Metadata