security-and-validation

Treat all client, command, file, network, and external-service input as untrusted until validated by the authoritative owner.

Validate

Check identity, ownership, permissions, target validity, range, state, quantity, bounds, rate limits, resource availability, and replay behavior. Recompute protected values on the server instead of trusting client displays.

Fail safely

Reject malformed or unauthorized requests without partial mutation. Avoid exception-driven control flow for expected rejection. Rate-limit repeated abuse and avoid revealing hidden state through distinct error messages.

Secrets

Never commit tokens, credentials, private keys, webhook secrets, or production connection strings. Use repository secrets and environment configuration. Avoid logging sensitive payloads.

Files and paths

Normalize and constrain user-controlled paths. Prevent traversal, arbitrary overwrite, shell interpolation, and unsafe temporary-file behavior. Use atomic replacement and explicit permissions where persistence requires it.

Verification

Test unauthorized access, malformed payloads, boundary values, replay, concurrent requests, stale targets, and partial failure.