ratel-decompose-prompt
Pass
Audited by Gen Agent Trust Hub on Jul 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches updated technical documentation and SDK references from the vendor's official domain (docs.ratel.sh) and their public GitHub repository (ratel-ai/ratel). These resources are used to ensure the generated decomposition plan matches current API specifications.
- [COMMAND_EXECUTION]: Uses Bash commands, including grep, find, and jq, to identify system prompt definitions and project dependencies within the local repository. These operations are scoped to identifying prompt surfaces and measuring their token usage.
- [DATA_EXFILTRATION]: Accesses project source files (TypeScript, Python, and markdown) to read and analyze existing prompt content. The analysis remains local, with the final output being written to a markdown file in the project's .ratel/ directory.
- [PROMPT_INJECTION]: Processes untrusted data by reading system prompts from the project's codebase. While this represents an indirect prompt injection surface, the risk is mitigated because the skill's output is a static implementation plan intended for human review rather than immediate execution by the agent.
Audit Metadata