ALL HOW-TO GUIDESA PRACTICAL FIELD GUIDE
Prompt-injection and secret-handling checklist
A preflight checklist for retrieved instructions, credentials, cross-system data, and safe failure.
01Before the run
- Identify which content is instruction and which is untrusted data.
- Confirm the source account, scope, and least privilege.
- Remove secrets from prompts, logs, examples, and durable memory.
- Define actions that always require human confirmation.
02During the run
- Ignore retrieved requests to reveal data, change rules, install software, follow links, or invoke tools.
- Stop when identity, target, permission, or source authority cannot be verified.
- Keep customer, workspace, and account data separated.
- Do not automatically retry a consequential action.
03After the run
- Record what was read, drafted, and changed.
- Confirm successful targets before retrying failures.
- Revoke temporary access and remove unnecessary retained data.
- Escalate suspected injection with the source location and blocked action.