Explore Prompts

Create a policy and technical controls for PII/secrets: detection, redaction, encryption, and safe storage. Include test cases and a plan to prevent secret leakage into training/evals.

Design data governance for user feedback and logs: consent, retention, minimization, access controls, and audit trails. Provide a policy and an implementation checklist.

Create a canary strategy: cohort selection, metrics, guardrails, and automatic rollback conditions. Include steps to prevent canary contamination and to interpret results statistically.

Design a rollback system: model version pinning, feature flags, staged rollouts, and a kill switch. Include operational procedures and testing for rollback reliability.

Create an incident response plan specific to AI: detection, containment, user comms, rollback, forensic logging, and post-incident retraining rules. Include severity levels and example incidents.

Define monitoring signals: policy violations, anomaly detection, tool misuse attempts, unusual output distributions, and drift. Provide alert thresholds, runbooks, and an on-call playbook.

Design a continuous red team program: scenarios, cadence, severity scoring, triage workflow, and how findings feed back into the improvement loop. Include a template for red-team reports.

Create a safety regression suite: prompt injection tests, data leakage tests, refusal/guardrail tests, and policy adherence checks. Include how to maintain and evolve the suite over time.

Design an offline sandbox environment for experimenting with improvements: isolated data, limited tools, no external side effects, and deterministic replay. Provide a checklist for containment.

Design an evaluation ladder for recursive improvement: unit tests, integration tests, simulation, canaries, and production monitoring. Provide pass/fail gates and minimum coverage targets.

Create a boundary model that separates: advisory outputs, drafts, and automated actions. Include criteria for graduating features from advice→action, and safety evidence required.

Define where humans must remain in the loop: high-impact actions, security-sensitive steps, and ambiguous decisions. Provide a decision taxonomy and UI/ops requirements for approval workflows.