The Agent Skills Directory

[SAFE]: The skill focuses on workflow organization and process verification. It includes explicit instructions to redact sensitive data during handoffs and ensures that external actions, such as publication or commits, require appropriate user authorization.
[INDIRECT_PROMPT_INJECTION]: The skill coordinates the processing of external data (project artifacts, logs, metrics, and primary sources). It mitigates potential injection risks by instructing the agent to distinguish evidence from assumptions and providing boundary instructions that restrict additional user input from authorizing unrelated external actions.
[EXTERNAL_DOWNLOADS]: The skill references guidelines and source revisions from well-known developers and organizations (including Matt Pocock, Addy Osmani, and the Cursor team). these references are documentation-focused and do not involve runtime remote code execution or automated downloads of untrusted scripts.

flow