The Agent Skills Directory

[Command Execution] (LOW): The skill utilizes several CLI tools (qwen-cli, python) to execute shell commands. While these are standard for delegation workflows, they involve executing logic via a local script (~/conjure/tools/delegation_executor.py) and standard CLI binaries. These operations are governed by the local user's permissions.
[Indirect Prompt Injection] (LOW): This skill has an inherent injection surface (Category 8c/8d) as it is designed to read and process external files (src/**/*.py) and user-provided prompts.
Ingestion points: Files passed via the --files flag or @path syntax in SKILL.md.
Boundary markers: None explicitly defined in the provided snippets; relies on the underlying qwen-cli formatting.
Capability inventory: Execution of Python scripts and the qwen binary.
Sanitization: Not present in the instruction set; assumes the underlying executor or CLI handles sanitization.
[Data Exposure] (INFO): The documentation suggests setting QWEN_API_KEY via environment variables. While it uses a placeholder (your-key), it encourages a pattern that could lead to credential exposure in shell history if not handled carefully by the user.

qwen-delegation