ometere123/the-skills-bible/ai-incident-response-engineer

AI Incident Response Engineer

Purpose

Contain AI-specific harm while preserving evidence, user protection, system integrity, and trustworthy recovery.

Use this skill when

• An AI system produces material harmful or incorrect outcomes.
• An agent takes unsafe or unauthorised action.
• Prompts, data, models, indexes, or tools may be compromised.
• A provider or model change causes unexpected behaviour.

Do not use this skill as the primary workflow when

• The issue is a routine low-impact output correction.
• The primary event is a general infrastructure outage.
• Authority to contain or investigate is missing.

Required inputs

• Incident signals and affected users.
• Exact AI system versions and traces.
• Models, prompts, retrieval, tools, memory, and policies.
• Data, privacy, security, and business impact.
• Containment, communication, and recovery capabilities.

Classify missing information as blocking, important, or optional. Inspect the actual product requirements, workflows, data, models, prompts, tools, permissions, evaluations, traces, incidents, policies, and production constraints before asking for information that can be verified directly.

Core operating principles

• Start from the decision, workflow, user outcome, or automation need rather than from a model or vendor.
• Use the simplest deterministic method that can satisfy the requirement before adding probabilistic components.
• Separate model capability, application logic, tools, memory, retrieval, policy, and human authority.
• Treat prompts, datasets, tools, schemas, model versions, policies, and evaluations as versioned production artefacts.
• Ground consequential outputs in authoritative evidence and expose uncertainty, provenance, and limitations.
• Constrain permissions, data access, actions, spending, recursion, and runtime before granting autonomy.
• Evaluate representative tasks, edge cases, adversarial inputs, failures, and affected user groups before release.
• Preserve privacy, security, accessibility, human agency, appeal, and safe failure throughout the lifecycle.
• Make model and agent behaviour observable without logging secrets, private reasoning, or unnecessary personal data.
• Design rollback, fallback, escalation, incident response, and decommissioning before production deployment.

Workflow

1. Declare and classify

• State observed AI behaviour, harm, affected tasks, users, time, severity, and uncertainty.
• Assign incident command and AI specialists.

2. Freeze evidence

• Preserve model, prompt, policy, dataset or index, tool set, traces, outputs, approvals, and provider metadata.
• Protect sensitive content.

3. Scope the incident

• Identify versions, cohorts, tenants, sources, tools, actions, regions, and time windows affected.
• Check correlated failures.

4. Contain

• Disable feature, model, tool, route, memory, source, automation, or provider; reduce autonomy; require human review; or roll back.

5. Protect users

• Stop harmful actions, notify support, correct records, provide appeal, reverse reversible effects, and prioritise vulnerable users.

6. Investigate cause

• Separate prompt change, model update, data or index issue, injection, poisoning, tool defect, policy gap, drift, and misuse.

7. Eradicate

• Remove compromised data, prompts, credentials, tools, memories, indexes, artefacts, and unsafe configurations.
• Rebuild trusted components.

8. Recover

• Deploy validated versions through constrained rollout, restore human gates, reconcile actions, and monitor for recurrence.

9. Communicate

• Provide factual internal, user, vendor, regulatory, and public updates through authorised owners without overstating attribution.

10. Learn and improve

• Create evaluation regressions, architecture changes, controls, monitoring, governance updates, and completed corrective actions.

Decision rules

Required outputs

• Incident declaration.
• Evidence preservation.
• Scope assessment.
• Containment record.
• User-protection actions.
• Cause analysis.
• Eradication.
• Recovery validation.
• Communication record.
• Corrective actions and regressions.

AI system quality gates

• Problem gate: The intended outcome and the reason to use AI are explicit.
• Data gate: Sources, rights, quality, lineage, privacy, and retention are controlled.
• Model gate: Model capability, limits, version, configuration, and fallback are documented.
• Grounding gate: Claims and actions use suitable evidence, tools, or deterministic checks.
• Authority gate: The model cannot exceed the permissions, spending, scope, or actions granted by policy.
• Evaluation gate: Representative, adversarial, failure, and subgroup evaluations meet approved thresholds.
• Human gate: High-impact, ambiguous, irreversible, or exceptional outcomes receive appropriate human review.
• Safety gate: Abuse, injection, privacy, harmful content, and unsafe action paths are tested and bounded.
• Observability gate: Inputs, outputs, tools, state, latency, cost, and outcomes are diagnosable with data minimisation.
• Lifecycle gate: Versioning, rollout, rollback, monitoring, incident response, and retirement are operational.

Evidence standard

For consequential AI conclusions, identify the exact model, model version, system prompt, tool set, retrieval index, dataset, policy, evaluation set, configuration, environment, and time window. Distinguish observed behaviour, measured estimates, supported inference, and unknowns. Do not claim capability, safety, fairness, or reliability from a few demonstrations.

Failure handling and escalation

Pause or escalate when:

• The system's intended authority, affected users, or acceptable error is unclear.
• Training, retrieval, or evaluation data lacks provenance, permission, or privacy controls.
• A model or agent may send messages, move money, change records, execute code, or access sensitive systems beyond approved scope.
• Prompt injection, tool compromise, data poisoning, identity confusion, or uncontrolled recursion is plausible.
• A high-impact outcome cannot be independently verified or appealed.
• Evaluation evidence is missing, stale, non-representative, or materially regresses.
• Residual safety, privacy, legal, financial, or operational risk exceeds delegated authority.

Report the exact blocker, verified state, affected outcome, safe options, and smallest required decision.

Common anti-patterns

• Deleting traces before investigation.
• Fixing a prompt while leaving broad tool authority.
• Treating harmful output as only a model issue.
• Continuing autonomous actions during uncertainty.
• No correction for affected records or users.
• Restoring the same poisoned index.
• Closing without evaluation regressions.

Completion checklist

• [ ] Exact artefacts and affected scope are known.
• [ ] Harmful authority is contained.
• [ ] Users and records are protected.
• [ ] Recovery uses validated trusted components.
• [ ] Communication is authorised.
• [ ] Incident lessons become durable evaluations and controls.

AI implementation record

# AI Implementation Record
 
## Outcome
[Task, user need, decision, or automation result]
 
## Scope and authority
[Users, systems, data, tools, permissions, budgets, and exclusions]
 
## Architecture
[Models, prompts, retrieval, memory, tools, policies, humans, and fallback]
 
## Data and provenance
[Sources, rights, lineage, privacy, retention, and quality]
 
## Evaluation
[Tasks, datasets, metrics, thresholds, adversarial cases, and subgroup results]
 
## Safety and security
[Injection, abuse, harmful output, tool constraints, privacy, and escalation]
 
## Deployment
[Versioning, routing, rollout, rollback, compatibility, and cost]
 
## Observability
[Traces, outcomes, latency, tokens, tool calls, incidents, and redaction]
 
## Residual risk
[Known limits, owner, acceptance, expiry, and user communication]
 
## Follow-up
[Monitoring, retraining, index refresh, policy updates, and decommission triggers]

Related skills

• ai-red-team-engineer
• ai-observability-engineer
• model-monitoring-drift-engineer
• security-incident-response-engineer

Maintenance rules

Review this skill when models, datasets, policies, tools, regulations, or workflows change; evaluations or incidents reveal a missing safeguard; the workflow becomes tied to one vendor; the file approaches 500 lines; or reusable schemas, evaluators, test sets, or guardrail policies would improve execution.