The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads the official Clerk CLI via standard package managers.
Evidence: SKILL.md mentions npx -y clerk@{{CLI_VERSION}}, bunx clerk@{{CLI_VERSION}}, pnpm dlx clerk@{{CLI_VERSION}}, and yarn dlx clerk@{{CLI_VERSION}}.
[COMMAND_EXECUTION]: The skill executes shell commands using the clerk binary.
Evidence: SKILL.md contains numerous examples of clerk commands like clerk api, clerk users list, and clerk config pull.
[CREDENTIALS_UNSAFE]: The skill handles sensitive API keys but follows industry standards for management.
Evidence: SKILL.md and references/auth.md recommend using CLERK_SECRET_KEY and CLERK_PLATFORM_API_KEY environment variables or the system keychain instead of hardcoding values.
[DATA_EXFILTRATION]: The skill performs network operations to Clerk's official Backend and Platform APIs.
Evidence: references/auth.md describes the use of https://api.clerk.dev/v1/ and https://api.clerk.com/v1/ for operations.
[SAFE]: The skill provides robust safety mechanisms for agent-based execution and handles data ingestion surfaces safely.
Evidence: SKILL.md and references/agent-mode.md document the use of --dry-run to preview mutations, the requirement of --yes for destructive calls in agent mode, and recommendations to save large API responses to local files for inspection with jq to avoid context flooding.

clerk-cli