The Agent Skills Directory

[COMMAND_EXECUTION]: The instructions direct the agent to perform high-privilege administrative tasks, such as rotating keys, renewing certificates, and unsealing services like HashiCorp Vault. It also describes procedures for modifying RBAC permissions and performing a 'force-unlock' on shared resources like Terraform states.
[PROMPT_INJECTION]: The skill includes instructions to ingest and act on untrusted user data regarding access barriers (SKILL.md). It attempts to mitigate injection risks through mandatory restatements of targets and explicit confirmation gates before taking action. Capabilities involve sensitive operations like key rotation and state unlocking (SKILL.md), but sanitization is limited to verification of the 'legitimate path'.
[SAFE]: The skill explicitly forbids the agent from cracking security controls, brute-forcing access, or removing compliance gates, ensuring that only authorized recovery mechanisms are utilized.

knock