Skip to content

security

Secret protection — NEVER/ALWAYS:

  • NEVER stage .env, .pem, .key, credentials.*, secrets.yml, SSH keys, or service account files
  • NEVER hardcode API keys, tokens, passwords, or connection strings in source code — use environment variables or gitignored config files
  • ALWAYS use .env.example / .env.template for documenting required variables (placeholders, no values)
  • ALWAYS check git status before committing to verify no sensitive files are staged

These four are yours because the hooks cover only part of the surface — privacy-guard.cjs (Read/Glob/Grep only), secret-guard.cjs (staging/pushing, content scan warn-only), both fail-open, and nothing scans a file at write time. Coverage matrix: docs/security.md.

When working with secrets — ask which env vars are needed → .env.example with placeholders → the user creates .env with real values → .env stays in .gitignore.

Shared credentials (used by more than one session) live in the studio vault, never per-session, never in a repo or chat: read with t1k vault get <name>; register, grant and the full convention are in the t1k-vault skill (module t1k-vault, required; doctor #96). Carve-out: a per-game Android keystore is committed at the Unity or Cocos project root; its passwords never are (docs/security.md).

Fail-open rationale and the hook coverage matrix: docs/security.md · docs/agent-security-boilerplate.md.