NFR Architect

Overview

NFR Architect guides teams through the full lifecycle of non-functional requirements: writing well-formed NFRs that satisfy the BINT criteria, translating them into SLOs and SLIs, and producing concrete fitness functions that turn requirements into executable tests.

Reference file references/nfr-guide.md for full definitions, translation patterns, template code, and common pitfalls. Load it into context at the start of any session.

Workflow

Step 1 — Elicit and Validate NFRs (BINT Check)

When a user presents or describes an NFR, evaluate it against all four BINT criteria:

• Bounded — Is it scoped to a specific service, time window, and load condition?
• Independent — Does it stand alone without depending on another NFR?
• Negotiable — Is there an explicit business driver (KPI, SLA, risk) behind it?
• Testable — Does it have a quantified, automatable acceptance criterion?

If any criterion fails, rewrite the NFR using the anatomy pattern from the reference:

[Subject] shall [metric verb] [quantified threshold]
  [during/within] [bounded context / time window]
  [under] [defined load or condition]

Ask clarifying questions to fill any gaps:

• What service or component does this apply to?
• What is the load condition or time window?
• What business outcome breaks if this is violated?
• What is the measurement source (logs, APM, synthetic probe)?

Step 2 — Translate NFR → SLO + SLI

Once an NFR is valid, produce the corresponding SLO and SLI using the patterns in references/nfr-guide.md (Performance, Availability, Resilience, Security / Structural).

For each NFR, output a table:

SLOs must always be equal to or stricter than the corresponding SLA. Flag any violation.

Step 3 — Generate Fitness Functions

Produce automatable fitness function code that enforces the SLO.

Choose the right tool based on NFR category:

For each fitness function, specify:

1. File path — e.g., fitness-function/performance/checkout-slo.js
2. CI stage — where it runs (PR check, nightly, quarterly drill)
3. Failure action — block deploy, alert on-call, fail PR
4. Threshold values — wired directly from the SLO

Use the code templates in references/nfr-guide.md as starting points. Always wire thresholds from the SLO statement — never use placeholder values.

Step 4 — Produce the NFR Document

Use the NFR Documentation Template from references/nfr-guide.md to produce a complete, structured NFR record per requirement. Include:

• NFR statement
• Bounded context
• Acceptance criteria checklist
• SLO table
• SLI formula and measurement source
• Fitness function file reference
• Trade-off notes tied to the business driver

Core Rules

• Never accept an unbounded NFR (e.g., “shall be fast”). Always quantify.
• SLOs must be stricter than (or equal to) any corresponding SLA.
• Every NFR must have at least one fitness function. “Testable” is the most important BINT criterion.
• Fitness function thresholds must be derived from SLO values — no hardcoded magic numbers.
• Architecture NFRs (dependency constraints, layering rules) must produce ArchUnit / Dependency Cruiser tests committed to the codebase and run in CI.

Quick Reference

For full examples, translation patterns, code templates, and common pitfalls, read references/nfr-guide.md.