The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external specifications to generate implementation plans, representing a surface for indirect prompt injection. (1) Ingestion points: User-provided task requirements and specifications. (2) Boundary markers: The skill does not implement delimiters to distinguish input data from its own instructions. (3) Capability inventory: The skill can write files to the docs directory, execute shell commands via generated tasks, and delegate actions to subagents. (4) Sanitization: No sanitization of user specifications is performed.
[DYNAMIC_EXECUTION]: The skill instructs the agent to generate Python source code and test scripts for runtime execution. Evidence: The Task Structure section provides templates for creating implementation and test files.
[COMMAND_EXECUTION]: The skill prompts the agent to construct and run specific shell commands. Evidence: The templates include explicit commands such as pytest with specified flags and file paths.

task-breakdown