claat-writer
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill downloads and executes a Go-based linting tool directly from the author's GitHub repository (github.com/nogikun/claat) to validate the generated manual.md file.
- [COMMAND_EXECUTION]: The skill utilizes the sed utility to perform text transformations on the documentation before performing prose checks.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the user's project environment to generate documentation, presenting a surface for indirect instructions. 1. Ingestion points: Project source code, README files, configuration files, logs, and external URLs. 2. Boundary markers: Absent. No explicit delimiters or instructions are provided to distinguish between document content and potential instructions embedded in the input files. 3. Capability inventory: Shell command execution (go run, sed) and file system access. 4. Sanitization: Absent. Content from ingested files is processed directly to generate manual content without escaping or validation.
Audit Metadata