The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external requirements to generate implementation plans. Maliciously crafted specifications could induce the agent to include harmful shell commands or code in the resulting plan file (docs/plans/*.md), which is then executed by downstream processes.
Ingestion points: Processes external "specs or requirements" as primary input.
Boundary markers: Absent. The instructions do not define delimiters or warnings for the agent to ignore embedded instructions within the input data.
Capability inventory: Generates shell commands (pytest, git) and Python code to be executed via superpowers:executing-plans or superpowers:subagent-driven-development.
Sanitization: Absent.
[DYNAMIC_EXECUTION]: The skill generates script fragments and shell command sequences within implementation plans intended for runtime execution by subagents. This pattern of generating code and commands for automated execution is a known attack surface.
[COMMAND_EXECUTION]: The skill instructs the agent to generate shell commands for test execution (pytest) and version control (git), which are intended to be executed in the environment.

writing-plans