The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the tsh and tctl command-line tools to query Teleport session recordings. It follows best practices by resolving binary paths and using the user's existing authenticated profile.
[INDIRECT_PROMPT_INJECTION]: The skill processes session summaries, resource labels, and usernames which are considered untrusted, attacker-controlled data. The instructions and a dedicated security reference file (references/SECURITY.md) explicitly command the agent to ignore any directives found in this data and provide validation rules (e.g., UUID checks for session IDs) to prevent command injection. Ingestion points: tctl recordings ls and tctl recordings search output (SKILL.md Step 2 and 3). Boundary markers: Explicit warnings in SKILL.md and references/SECURITY.md tell the agent to treat all tctl output as untrusted. Capability inventory: The skill can execute local binaries (tsh, tctl) and perform network requests (curl). Sanitization: The security rules mandate validating session IDs as UUIDs and strictly prohibit interpolating free-text fields into shell arguments.
[EXTERNAL_DOWNLOADS]: Fetches cluster configuration from the Teleport proxy (/web/config.js) via curl to determine if advanced features like session search are supported. This is a read-only operation targeting the user's own authenticated infrastructure.

teleport-session-review