The Agent Skills Directory

[SAFE]: The skill is a documentation and advisory tool that does not perform code execution, network operations, or sensitive file system modifications. It is designed to guide an agent through architectural discovery and auditing without implementing or running code.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external agent specifications, source code, and execution traces provided by the user. While this creates a theoretical surface for indirect prompt injection, the skill includes explicit instructions to mitigate risk, such as forbidding the execution of applications during an audit and requiring human-in-the-loop confirmation for significant actions.
[REMOTE_CODE_EXECUTION]: The skill does not contain any remote code execution or automated script downloads. It explicitly instructs the agent to separate architecture design from implementation and states that it 'does not authorize writing code, changing agent settings, publishing, or initiating external actions'.

agent-architecture