The Agent Skills Directory

[SAFE]: The skill serves as a legitimate software design assistant. It implements a structured workflow that emphasizes scope control and verification without employing any suspicious or malicious patterns.- [PROMPT_INJECTION]: The skill explores the codebase and reads documentation to build context. Although this creates a surface for indirect prompt injection, the risk is mitigated by the skill's interactive design. The agent is instructed to ask one question at a time and confirm every decision explicitly with the user, which prevents the automatic or silent execution of potentially malicious instructions embedded in the codebase.- [COMMAND_EXECUTION]: The skill interacts with the local filesystem to create a 'specs' directory and save markdown files. These actions are within the expected scope of a specification tool and occur only after the content has been finalized through user interaction.

specify