SABERISTIC / PRODUCTION READINESS CHECK
Find the gate before it becomes the incident. Answer 20 controlled architecture questions. Get a defensible readiness level, hard blockers, unknowns, a 48-hour plan, and a two-week production plan.
Scored by explicit controls. Explained by AI when the privacy-safe route is available.
01 / CHOOSE A PROFILE
What are you building? The profile supplies context only. It cannot change an answer, score, blocker, or next step.
01 AI-enabled SaaS A web product using accounts, a managed database, payments, or an AI provider. 02 Agent workflow Tools, retrieval, memory, and material actions that need an approval boundary. 03 Payments product A product involving money movement, payment events, wallets, or reconciliation. 04 Something else Start without assumptions about the architecture, stage, or business model.
02 / DECLARED CONTROLS
Section 1 of 5 · 0 of 20 answered 1 of 5 Stage and architecture Current reach, consequence, environment boundaries, and critical dependencies.
01 Who can reach the system today? Choose the boundary that is reachable now, not the audience planned later. Internal, without accounts Only a known internal group can reach it, and it has no user accounts. Internal, with accounts Only a known internal group can reach it, and users sign in. Public, without accounts Anyone can reach it, but it does not expose user accounts. Public, with accounts Anyone can reach it and it exposes user or operator accounts. Unknown The deployed reachability or account boundary has not been verified.
Question 1 in this section. 02 What happens if the most important journey fails or returns a wrong result? Classify the highest credible consequence, even if it is unlikely. Low and reversible The impact is limited to an easily repeated demo or informational experience. Contained customer impact A bounded group is inconvenienced, with no material data or financial consequence. High business or data impact Customers, operations, revenue, or important data can be materially affected. Critical or irreversible impact A failure can create safety, legal, irreversible financial, or severe confidentiality harm. Unknown No one has classified the consequence of a wrong result or outage.
Question 2 in this section. 03 How is the current release separated from development and testing? Consider credentials, data stores, deployment permissions, and configuration together. Not deployed There is no production deployment yet. Fully isolated production Production has separate credentials, data, configuration, and deployment permissions. Partially separated Some production boundaries are separate, but at least one material resource is shared. Material production state is shared Development or tests can touch live credentials, data, or deployment state. Unknown The environment boundary has not been reviewed.
Question 3 in this section. 04 How are critical dependency failures bounded? Include databases, identity, payment, model, queue, storage, email, and external API providers. Inventoried and bounded Critical dependencies are inventoried with timeouts, bounded retries, and safe failure behavior. Known but partly bounded The dependencies are known, but some timeouts, retry limits, or degradation paths are missing. Unbounded or unmanaged Critical calls can wait, retry, cascade, or fail without an explicit boundary. Unknown Critical dependencies or their failure behavior have not been identified.
Question 4 in this section. Continue
METHOD AND AI BOUNDARY
The model never owns the result. Versioned application rules calculate the score, level, completeness, blocker set, and next step before any model call. OpenRouter can only explain and order approved actions. Invalid, unavailable, over-budget, or non-private model routing returns the same complete deterministic report.
◇ This is directional engineering guidance, not a security audit, compliance assessment, certification, or inspection of your system.
Read the privacy boundary 