The Agent Skills Directory

[SAFE]: The skill is entirely instructional, providing a framework for managing complex development tasks through document phases (Requirements, Design, Tasks).
[SAFE]: It includes explicit security-positive instructions to only use local relative paths for sibling skills and strictly forbids fetching remote skill or protocol markdown via HTTP.
[INDIRECT_PROMPT_INJECTION]: The skill involves processing user requests into structured documentation. It manages the risk of accidental execution or misinterpretation by requiring mandatory user confirmation checkpoints between each phase of the workflow.

spec-workflow