The Agent Skills Directory

[COMMAND_EXECUTION]: The skill interfaces with the tctl CLI through a strictly defined allowlist of commands. It utilizes robust single-quote shell-escaping (e.g., converting ' to ''') for all user-supplied variables to mitigate command injection risks.
[INDIRECT_PROMPT_INJECTION]: The instructions proactively defend against indirect prompt injection by teaching the agent to recognize and flag malicious instructions (e.g., 'ignore previous instructions') embedded in resource metadata, such as Teleport list descriptions, rather than following them.
[SAFE]: A strict 'human-in-the-loop' model is followed for all data-modifying actions. The skill is designed to present a clear final draft of any command intent and must wait for explicit user confirmation before execution.
[SAFE]: Temporary directories used for capturing tool output are handled securely using mktemp -d for isolation, with mandatory cleanup using rm -rf restricted to the generated path immediately upon completion.
[NO_CODE]: This skill consists entirely of markdown instructions and does not bundle any executable scripts or binary dependencies, relying instead on pre-installed and authenticated system utilities for its operations.

teleport-acl-lifecycle