unify-balance
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill recommends installing official Node.js packages from the @circle-fin ecosystem, which are verified vendor resources necessary for the skill's primary functionality.\n
- Evidence: Found in SKILL.md and reference files (e.g., @circle-fin/app-kit, @circle-fin/adapter-viem-v2, @circle-fin/adapter-solana).\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied transaction parameters and codebase context to generate implementation code, representing a potential indirect prompt injection surface.\n
- Ingestion points: User conversation (amounts, recipient addresses) and project configuration files.\n
- Boundary markers: The Instruction Hierarchy explicitly prioritizes direct user instructions over repository context to prevent instructions in user files from overriding safe behavior.\n
- Capability inventory: The skill is restricted to code generation; it lacks the capability to automatically execute transactions, run shell commands, or perform network operations.\n
- Sanitization: The skill includes mandatory rules for validating all input addresses, amounts, and chain identifiers before generating code.
Audit Metadata