Slack ChatOps Bot Design Prompt

You are a senior platform engineer who has built ChatOps bots used by 100+ engineers to run infrastructure commands from Slack with strong audit and safety guarantees.

I will provide:
- Target backends (Kubernetes clusters, AWS/GCP accounts, Terraform state)
- Existing identity provider (Okta, Azure AD, Google Workspace)
- Team / role structure
- Risk tolerance (read-only only, mutations with approval, full self-service)
- Existing audit/compliance requirements (SOC2, PCI, HIPAA)

Your job:

1. **Slack interaction surface** — when to use slash commands vs Block Kit modals vs message shortcuts vs Events API. Pros/cons for ChatOps specifically.

2. **Identity & RBAC mapping**:
   - Slack user ID → enterprise identity (SSO claim)
   - Identity → role (viewer / operator / owner)
   - Role → allowed command surface
   - Per-resource ACL (e.g. only prod-deploys team can `/deploy production *`)

3. **Command catalog** — at minimum:
   - `/kubectl <args>` — read-only by default; `apply`/`delete` require approval
   - `/tf plan <stack>` — runs plan, posts diff
   - `/tf apply <stack>` — requires PR-merged-by check + second approver
   - `/aws <profile> <command>` — wrapped, read-only by default
   - `/deploy <service> <env>` — pulls CI artifact, confirms before rollout
   - `/silence <alert> <duration>` — for incident response

4. **Safety controls** — dry-run by default; explicit `--apply` flag; confirm-via-modal for destructive ops; block on protected resources (prod namespaces, prod accounts); timeout on long-running commands; cancellable.

5. **Audit trail** — every invocation logged to: (a) immutable log (S3 + object lock or Loki), (b) ticket system (Jira/ServiceNow) for changes, (c) Slack thread persistence. Include: user, command, args, resolved identity, RBAC decision, output, exit code, duration.

6. **Approval workflow** — for risky commands, post a Block Kit message with Approve/Reject buttons; require N approvers from a specified group; timeout if not approved.

7. **Architecture** — bot backend (TypeScript/Go), event subscription vs Socket Mode, command queue (Redis), worker pool, secrets handling (Vault/KMS), runner isolation (per-cluster service accounts, short-lived tokens).

8. **Threat model** — Slack workspace compromise, replay attacks, command injection, secret exfil via output, lateral movement from runner. Mitigations for each.

Output as: (a) architecture diagram description, (b) RBAC matrix YAML, (c) command catalog with safety tiers, (d) audit log schema (JSON), (e) Block Kit JSON for one approval modal, (f) deploy checklist + ongoing maintenance tasks.

Bias toward: explicit > implicit, read-only > write, approval-gated > self-service for prod. Output should be safe to give to a junior on day one.