Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
DeepSeek
Muse+22 moreClaude CodeCursorCodex CLIDeepSeekMuse
Dots
Manus
Grok
Grok Bot+18 agentsEnterprise Automation Architect — Production Workflow Design & Reliability
The problem Teams “automate the happy path” and discover duplicates, missed revokes, and silent API failures in production.
$18
Enterprise Automation Architect — Production Workflow Design & Reliability
Example session with this skill installed
Process: CampusAccess — after-hours building access exceptions
Manual today
- ServiceNow request (building, rooms, window, reason)
- Clerk checks Workday employment active
- Clerk checks LMS safety training complete
- Clerk creates temporary badge grant in access-control system
- Slack #campus-security + email requester
- BSL-2 / lab after 10pm needs lab director signature (PDF today)
- Morning revoke often forgotten → orphan grants
Volume: ~80/week, ~200/week at semester start
SLA: 4 business hours for standard buildings
Constraints: reuse Temporal; no secrets in repo; no orphan grants after window
Ask: full production automation architecture (design mode). Do not execute.
- Read your context and instructions
- Compiled the enterprise automation architect
Enterprise Automation Architecture
1. Decision Summary
| Field | Value |
|---|---|
| Process | CampusAccess after-hours access exceptions |
| Mode | Design |
| Paradigm | event-driven (ServiceNow) |
| Scale band | medium at semester peaks |
| Build vs buy | Orchestrate on Temporal + modular adapters |
| Maturity | NEEDS-HARDENING |
2. Human-automation boundary (excerpt)
| Step | Verdict |
|---|---|
| Workday + LMS checks | AUTOMATE |
| Standard building grant | AUTOMATE (after checks) |
| Lab / BSL-2 after 10pm | KEEP-MANUAL path with director APPROVE |
| Revoke after window | AUTOMATE (compliance) |
| Policy waivers | KEEP-MANUAL |
3–4. Paradigm & build recommendation
Event-driven async workflow (not nightly batch). Reuse campus Temporal for
durability, approval waits, and revoke timers. Thin adapters for ServiceNow,
Workday, LMS, access control, Slack/email — not a second iPaaS.
5–6. Contracts & reliability
- Grant create: effectively-once via idempotency_key = ticket_id
- Webhook: at-least-once with dedupe
- Notify: at-most-once (never block grant on Slack failure)
- Exception tree: inactive/untrained → auto-deny; access auth fail → escalate;
lab after 22:00 → park for APPROVE; revoke fail → escalate + orphan report
7–8. Security & HITL
Threats: stolen creds, over-privileged bot, webhook spoofing, PII in logs.
Secrets manager only. Least-privilege grant/revoke role.
HITL: lab-after-10pm director APPROVE; standard path AUTO when policy passes.
9–10. Observability & rollout
Metrics: auto-approve rate, time-to-grant, orphan_grants_count, revoke failures.
SLO: 4h standard decisions; zero orphans older than 24h.
Rollout: Shadow (simulate) → Canary (one non-lab building) → Full (labs stay APPROVE).
11–14. ROI, testing, runbook, maturity path
Labor assumption ~20 clerk hours/week; payback often 1–2 semesters if validated.
Tests: policy, idempotency, revoke timer, chaos on access API.
Runbook: detect → triage by ticket_id → recover → communicate.
To PRODUCTION-READY: confirm OpenAPI maps, orphan scanner, director UX, shadow period.
15. JSON
Full Automation Architecture JSON (boundary, adapters, steps, threats, slo, rollout,
cost_model, testing, runbook, assumptions).
Connects securely to your tools. The creator never sees your data.
About this skill
The problem
- Teams “automate the happy path” and discover duplicates, missed revokes, and silent API failures in production.
- Not every step should be automated — high-judgment and high-liability steps need human-in-the-loop design up front.
- Build-vs-buy and orchestrator choice are often made by habit, not by volume, durability, and ops ownership.
- Leadership asks for ROI and SLA accountability, but the design never defined metrics or a safe rollout.
What it does
- Classifies the automation paradigm (batch, scheduled, event-driven, streaming, human-triggered) and scales the design to real volume.
- Draws a human-automation boundary: AUTOMATE, ASSIST, or KEEP-MANUAL per step with reasoning.
- Recommends build vs buy vs orchestrate (e.g. Temporal, Airflow, n8n, SaaS) with a clear rationale.
- Designs modular adapters with data contracts and delivery guarantees (at-most / at-least / effectively-once).
- Specifies idempotency, retries, rate limits, and an exception decision tree (retry / escalate / skip / compensate / park).
- Builds a security threat model, secrets plan, and HITL gates (AUTO / CONFIRM / APPROVE).
- Defines observability + business metrics + SLA/SLOs, a shadow → canary → full rollout, cost/ROI, testing, runbook, and a maturity path to production.
Frameworks & tools
- Orchestrators and integration platforms commonly used in enterprises (Temporal, Airflow, n8n, iPaaS) plus custom adapters.
- Works across ERP, ITSM, HRIS, LMS, messaging, access control, and other SaaS/API systems you name.
- Compatible with Claude Code, Cursor, Codex CLI, OpenClaw, and other SKILL.md agents — no live execution required for design.
Why this beats prompting it yourself
- Forces the hard questions: what not to automate, what needs approval, how duplicates are prevented.
- Typed delivery guarantees and exception trees instead of “add a retry.”
- Security and PII logging constraints designed with the workflow, not bolted on later.
- Rollout stages with exit criteria and rollback — not a big-bang cutover.
- ROI and SLOs stated with explicit assumptions, not invented precision.
- Maturity verdict with a concrete path (e.g. NEEDS-HARDENING → PRODUCTION-READY).
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
• Design and plan only — not running code or deployed infrastructure
• Vendor field maps and dollar costs are assumptions unless you supply them
• Not for pure desktop UI RPA as the primary approach
• Not a runtime reliability or SOC2 certification
• Does not execute the automation or write production data
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 21 days ago
Permissions
This skill only analyzes the process and systems you describe and drafts an architecture. It does not run jobs, call APIs, change data, or store secrets. Grant Read only; leave Shell, Write, Browser, Network, and Env off.