The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to include specific directives in its output documents (implementation plans) to trigger subsequent agent actions, such as 'REQUIRED SUB-SKILL: Use superpowers:executing-plans'. This facilitates a multi-step task chain where output from one session influences the instructions of the next.
Ingestion points: The skill processes user specifications to generate Implementation Plans in the docs/plans/ directory.
Boundary markers: No specific delimiters are enforced in the output templates to distinguish between data and instructions for downstream agents.
Capability inventory: The generated plans include shell commands for testing (pytest) and version control (git), along with Python code blocks.
Sanitization: No explicit sanitization or escaping of the user-provided requirements is described before interpolation into the plan document.
[COMMAND_EXECUTION]: The skill provides templates for the agent to generate shell commands for routine development tasks, specifically pytest for running unit tests and git for committing code changes.
[DYNAMIC_EXECUTION]: The skill generates implementation plans containing Python code snippets and shell commands meant for later execution. These follow standard templates for Test-Driven Development (TDD) and do not incorporate untrusted external sources during the generation process.

writing-plans