seed-verify-lynx-component
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill utilizes the agent-lynx evaluate command to execute arbitrary JavaScript code within the Lynx runtime environment. While this is an intended feature for inspection, it represents a high-capability dynamic execution sink.
- [COMMAND_EXECUTION]: The skill involves the execution of various CLI tools and shell-based logic for concurrency management, such as creating atomic lock directories in /tmp and using node for string encoding tasks.
- [EXTERNAL_DOWNLOADS]: The documentation suggests using bunx agent-lynx, which downloads and executes the agent-lynx package from an external package registry at runtime.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to read and verify untrusted external content, including component source code, MDX documentation, and runtime console logs, which could serve as vectors for indirect prompt injection. Mandatory Evidence Chain: 1. Ingestion points: Reads package exports, MDX examples, and runtime console/layout data (SKILL.md, references/verification.md). 2. Boundary markers: No specific delimiters or instructions to ignore embedded commands are provided for these data sources. 3. Capability inventory: Includes powerful tools such as agent-lynx (evaluate, snapshot), portless, and bun. 4. Sanitization: No sanitization or input validation of the external content is observed in the provided instructions.
Audit Metadata