The Agent Skills Directory

[SAFE]: The skill implements a procedural guardrail via the '' instruction, which strictly prohibits calling implementation tools or writing code until the user has reviewed and approved a design specification. This minimizes the risk of autonomous, unintended actions.
[COMMAND_EXECUTION]: The skill performs file system writes to a project's docs/tmp/ directory and executes git commits to store design documents. These actions are within the expected scope of a development tool and are necessary for the skill's primary function of requirements documentation.
[PROMPT_INJECTION]: The skill includes instructions to explore project context by reading files, documentation, and commit history. While reading untrusted project data is a surface for indirect prompt injection, it is the intended primary purpose of the skill in a developer context. The risk is mitigated by the skill's focus on gathering requirements through a structured dialogue with the user and the human-in-the-loop checkpoint during the design review phase.

brainstorming