The Agent Skills Directory

[DATA_EXFILTRATION]: The skill contains logic to load the CLERK_SECRET_KEY from .env and .env.local files.\n
Evidence: The _clerk_load_env function in scripts/_lib.sh walks up the directory tree to find and parse Clerk-prefixed environment variables.\n
Context: This is used solely for authenticating Backend API requests and is handled by a custom parser that prevents shell variable leakage.\n- [EXTERNAL_DOWNLOADS]: The skill downloads and executes the official Clerk CLI at runtime.\n
Evidence: bunx clerk@latest --mode agent is the primary method of execution mentioned in SKILL.md and references/commands.md.\n
Context: This targets the official clerk package on the npm registry, which is a standard and expected dependency for this skill.\n- [COMMAND_EXECUTION]: The skill makes extensive use of subshells to run CLI tools and network requests.\n
Evidence: Numerous scripts (e.g., scripts/clerk-api.sh, scripts/clerk-users.sh) invoke curl and bunx to perform management tasks.\n
Context: These operations are authorized by the user's explicit request and are necessary for the skill's stated purpose.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes external data that could potentially contain malicious instructions.\n
Evidence: API responses and user-provided metadata are handled by the scripts.\n
Mitigation: The skill uses jq to safely build JSON payloads, effectively preventing injection of malicious content into structured API requests.\n- [REMOTE_CODE_EXECUTION]: A preflight script references a common remote installation pattern.\n
Evidence: scripts/clerk-ensure.sh contains the string curl -fsSL https://bun.sh/install | bash.\n
Context: This is a diagnostic message suggesting how the user can install a missing dependency; it is not executed by the skill itself.

clerk-cli