Write an API Error Handling Playbook

Create an API error handling playbook for this service:

API style: [REST / GRAPHQL / RPC]
Language / framework: [STACK]
Audience of errors: [CLIENT APPS / PARTNERS / INTERNAL]
Existing status code conventions: [CONVENTIONS OR NONE]
Logging / observability stack: [TOOLS]
Common failure modes: [TIMEOUTS, VALIDATION, AUTH, DEPENDENCIES]

Requirements:
1. Define when to use 4xx vs 5xx and how to map domain errors.
2. Specify a stable error payload shape (code, message, details, request_id).
3. Include retry guidance for clients (idempotency, backoff, what not to retry).
4. Cover logging fields and what must never be logged (secrets, PII).
5. Give copy-paste examples for validation, auth, not found, and dependency failure.

Deliverables:
1. Error taxonomy table.
2. Example error JSON for four common cases.
3. A short implementation checklist for PR reviewers.