The Agent Skills Directory

[PROMPT_INJECTION] (SAFE): No patterns of instruction overrides, role-play jailbreaks, or safety bypass attempts were detected. The language is purely instructional and focused on CMS development.
[DATA_EXFILTRATION] (SAFE): No sensitive file paths (e.g., .ssh, .aws) or unauthorized network calls are present. Examples use environment variables for secrets, following security best practices.
[REMOTE_CODE_EXECUTION] (SAFE): No instances of piping remote content to a shell or dynamic execution of untrusted code. Command examples (npx, pnpm) are for developer project initialization.
[CREDENTIALS_UNSAFE] (SAFE): No hardcoded API keys or secrets. The config examples correctly reference process.env.PAYLOAD_SECRET and process.env.DATABASE_URL.
[OBFUSCATION] (SAFE): The content is entirely human-readable with no encoded strings (Base64), zero-width characters, or homoglyphs.

payload-cms