The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external specifications or requirements provided by the user to generate detailed implementation plans. While this is a core functional requirement, it creates an attack surface where malicious instructions in the input requirements could be incorporated into the resulting plan files.
Ingestion points: Input specifications or requirements for multi-step tasks.
Boundary markers: None specified in the instructions to separate user input from generated plan structure.
Capability inventory: Generates files in docs/plans/ that include shell commands and Python code.
Sanitization: No explicit sanitization of input data is defined before interpolation into the plan template.
[DYNAMIC_EXECUTION]: The skill generates Markdown documents containing Python source code and Bash command sequences (such as pytest and git commit). These plans are explicitly designed for downstream execution by specialized sub-skills like superpowers:executing-plans.
Evidence: The task structure template includes blocks for pytest command execution and automated git commits, facilitating a script-like execution flow for the agent.

writing-plans