The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to invoke the local qwen command-line interface. It provides specific guidance on using flags like --approval-mode (e.g., plan, default, auto_edit) and --model to manage the tool's level of autonomy and behavior during code generation or analysis.
[INDIRECT_PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by ingesting project context and user-supplied objectives to be processed by an external AI model. While the skill instructs the agent to use a prompt template and validate the output before implementation, malicious instructions contained within project files could potentially influence the model's suggestions. Evidence chain:
Ingestion points: User requests and project files identified in Step 1 and 2.
Boundary markers: A prompt template is provided in the instructions to structure the delegation.
Capability inventory: The skill utilizes Bash, Read, and Write tools.
Sanitization: The instructions mandate user confirmation (Step 5.2) and manual result validation before any code changes are applied.

qwen-coder