coding-protocol
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the processing of external repository data, creating a potential attack surface.
- Ingestion points: The agent reads code, tests, documentation, and runtime output as defined in SKILL.md.
- Boundary markers: The protocol directs the agent to resolve ambiguity, separate observations from assumptions, and surface factual mismatches before proceeding (SKILL.md, Section 3).
- Capability inventory: The agent is authorized to perform workspace mutations, execute local commands, and perform network operations (SKILL.md, Sections 4, 6, and 7).
- Sanitization: No explicit sanitization or filtering of processed repository content is implemented.
- [EXTERNAL_DOWNLOADS]: The skill provides an informational reference to a public post on X (a well-known social media platform) by Andrej Karpathy to provide the rationale behind the protocol's rules.
Audit Metadata