The Agent Skills Directory

[SAFE]: The skill functions as a security manifest that explicitly defines which commands are safe to auto-run and which require explicit user confirmation, ensuring human-in-the-loop for sensitive actions like identity management and data deletion.- [SAFE]: It implements proactive defenses against indirect prompt injection by instructing the agent to treat incoming message content as data rather than instructions.- [SAFE]: The rules provide clear guidance on the protection of sensitive information, prohibiting the exposure of private keys, tokens, or full identifiers in agent responses.- [SAFE]: The skill mandates the use of dry-run flags for side-effectful commands, reducing the risk of accidental or unintended state changes.

awiki-shared