Facet
DeutschInstall

FOR TEAMS USING SPEC-DRIVEN DEVELOPMENT

Already using specs? Check which connection is missing.

Spec-driven development and FACET can complement each other. What matters is whether requirements, implementation and evidence remain discoverable through later changes.

SPECspecs/discount-codes.md
## Discount codes- A valid code lowers the total once.- Two codes don't stack;  the higher one wins.- The discount is capped at 50 %.

the same rule, anchored in code after implementation

RECORDcheckout/pricing/discount-codes.yaml
contract:  scenarios:    - id: codes-do-not-stack      given: two valid codes      then: only the higher one counts
CODEsrc/checkout/discount.ts
// @facet: checkout/pricing/discount-codesexport function applyCodes(order, codes) { … }

Illustrative example

TWO POINTS IN TIME

The spec settles the before. FACET keeps the after.

BEFORE IMPLEMENTATIONSpec-driven Development

Clarify intent

Describe requirements before an agent writes code.

AFTER EVERY CHANGEFACET

Keep the connections

Connect rules with code, tests and knowledge — including existing products.

IN DETAIL

Four questions, two answers.

Scroll the table sideways to see spec-driven development and FACET side by side.

FACET and spec-driven development
QuestionSpec-driven developmentFACET
How does an idea become software?A specification defines the desired outcome; plans and tasks organize implementation.Feature records capture purpose, rules and scenarios. FACET does not replace implementation planning.
What happens after the first implementation?Depending on the method, specs evolve or follow other persistence models. Existing systems and iterative changes are also part of SDD.Code anchors connect records to implementation; configured checks flag changes for review.
Where is the evidence?Tests and acceptance evidence may be linked, depending on the process and tools used.Tests are mapped to scenarios. Available results and missing evidence can be viewed per feature.
Who reuses the knowledge?A specification can be a lasting shared reference.Atlas and generated help make recorded feature knowledge accessible to more readers.

For context: GitHub Spec Kit describes SDD including evolution of existing projects. This table compares typical tasks, not every SDD tool. GitHub Spec Kit

TOGETHER, NOT EITHER-OR

One workflow. One authoritative source.

Bring confirmed scenarios from your spec into suitable records. Anchor the implementation actually built, link real tests and review subsequent changes against the same rules. Explicitly decide which source is authoritative so you do not create two conflicting versions of the truth.

  1. SPECConfirm scenarios with the business
  2. RECORDCarry confirmed scenarios over
  3. ANCHORAnchor what was actually built
  4. EVIDENCEMap real tests to the scenarios

WHEN YOU DON'T NEED FACET

Your process already connects it all? Then skip the extra layer.

If your SDD process already connects rules, code, tests and ongoing updates reliably, FACET may create duplicate maintenance. Identify a concrete gap before introducing another layer.