The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external 'specs' or 'requirements' documents which serve as an ingestion point for untrusted data that could influence the agent's behavior.
Ingestion points: The skill takes a 'spec or requirements for a multi-step task' as its primary input (SKILL.md).
Boundary markers: There are no explicit instructions to the agent to disregard instructions embedded within the provided spec files.
Capability inventory: The skill defines file creation/modification paths, generates git commit commands, and produces shell commands for test execution (SKILL.md).
Sanitization: No automated sanitization of external text is defined; however, the skill mandates a 'Self-Review' checklist and provides a 'Plan Document Reviewer Prompt' as a manual/automated verification stage.
[DYNAMIC_EXECUTION]: The skill generates implementation plans that contain specific shell commands and code blocks intended for later execution by the user or a subagent.
Evidence: The 'Task Structure' section in SKILL.md provides templates for generating pytest and git commands, as well as Python code snippets, based on the content of the project specification.

writing-plans