qwen-dispatch

Warn

Audited by Gen Agent Trust Hub on Apr 18, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill defines an automatic pipeline mode (qwen-dispatch auto) described as 'Think + plan + execute'. This involves the execution of code or sequences of actions generated by an LLM at runtime.
  • [COMMAND_EXECUTION]: The skill provides various worker commands such as fix, scaffold, and refactor which are intended to modify the local filesystem and project structure based on model output.
  • [PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from a user's codebase (e.g., through the review, explore, and fix commands). This creates a surface for indirect prompt injection.
  • Ingestion points: Reads files from the workspace directory provided via the --cwd flag (e.g., SKILL.md).
  • Boundary markers: The instructions do not specify any markers or delimiters to isolate untrusted code content from the agent's instructions.
  • Capability inventory: The skill utilizes the qwen-dispatch CLI tool which has the capability to read files, modify the filesystem, and execute generated plans.
  • Sanitization: There is no mention of sanitization, validation, or escaping of the code content before it is processed by the model.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 18, 2026, 12:07 PM
Security Audit — agent-trust-hub — qwen-dispatch