- Home
- Skills
- APIs & Backend
- Architecture Integrity Reviewer
Works with the AI Agents you already use
Architecture Integrity Reviewer
Audit service boundaries, data flows, and deployment designs for hidden coupling, failure risk, and change cost before you build it.
Free
Architecture Integrity Reviewer
Example session with this skill installed
We're proposing a new order service that will share the existing customer PostgreSQL database and write directly into the inventory tables to keep the migration fast. Checkout currently calls the payment provider synchronously. The team is six engineers with one part-time SRE, and we ship daily. Can you review this design for architectural integrity — boundaries, data consistency, failure modes — compare a minimal option against a target option, and produce a decision record plus a validation plan we can bring to Friday's design review? Context docs are in docs/arch/.
- Read your context and instructions
- Compiled the architecture integrity reviewer
Summary
Order-service proposal reviewed: the shared customer database and direct writes into inventory tables create hidden coupling and an unowned failure path. Approved with conditions — establish data ownership and put inventory mutations behind an explicit contract before build starts.
Findings
| Area | Finding | Risk | Recommended action |
|---|---|---|---|
| Boundary | Order service writes inventory tables directly | High | Expose inventory via versioned API with idempotency keys |
| Data | Customer DB shared by two services | High | Split schema by domain; add outbox + reconciliation job |
| Resilience | Synchronous payment call inside checkout | Medium | Move to event stream with dead-letter handling |
| Observability | No queue-depth or retry telemetry planned | Low | Emit backlog metrics tied to checkout success rate |
Decision record (ADR-014)
Status: proposed. Chosen: modular monolith with domain-scoped schemas; extract the order service once the read/write split proves stable. Revisit if write volume doubles or a second team consumes order events.
Next steps
- Block new cross-service table access with lint/CI rules this sprint
- Add contract tests for the inventory API before any consumer ships
- Run a payment-outage chaos drill; define rollback as checkout success dropping below 99.5%
Connects securely to your tools. The creator never sees your data.
About this skill
The problem
System designs often suffer from hidden coupling, unmanaged failure modes, and high change costs that only surface after deployment. Developers struggle to objectively weigh architectural tradeoffs against team topology and operational reality.
What it does
- Identifies boundary erosion and integration leaks in microservices or modular monoliths.
- Traces failure paths including retries, idempotency, and cascading outages across service boundaries.
- Evaluates data consistency models, from strong transactional integrity to eventual consistency patterns.
- Produces formal Architecture Decision Records (ADR) with explicit tradeoffs, validation plans, and rollback conditions.
Why this beats prompting it yourself
This skill moves beyond generic design patterns to focus on the economics of change and operational health. It enforces a rigorous multi-step workflow that catches high-risk triggers, such as shared database integration or synchronous fan-out, which standard LLM prompts often overlook.
Use cases
- Reviewing a service decomposition plan to ensure independent deployability and clear ownership.
- Auditing event-driven systems for schema governance, message ordering, and retry safety.
- Comparing a modular monolith against a microservices transition based on team size and scaling needs.
- Assessing multi-region deployment designs for data residency and disaster recovery risks.
Known limitations
Does not perform local syntax or unit-test reviews. Requires clear business constraints and quality attributes to provide meaningful architectural evaluation.
The review is only as good as the context provided — it cannot measure live traffic, run load tests, or inspect production infrastructure. Regulated teams should still route the resulting decision record through their formal architecture board.
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 any payment.
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
3 uses
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 2 months ago
- Free with an Agensi account
Permissions
File Scopes
Read-only access to design docs, ADRs, and config files in the repository so the review is grounded in the actual system rather than just the proposal.