Synthesis contract

Report

Report is the synthesis capstone: fetch one URL, run score + drift + readiness in parallel, and produce a unified design-intelligence report with a single composite grade. One input, one output, one grade — the most shareable surface in the dynasty. Where /score answers "how good is this design", /drift answers "is AI breaking it", and /readiness answers "can agents use it", /report answers all three at once.

Version 0.1.0 · provisional · machine export

Source authority

Primary source

Designesy design intelligence research 2026-08-02

Composition

Weighted synthesis of three independent Designesy engines — /score (40-check audit), /drift (12-check drift radar), /readiness (10-check AI readiness). Each engine is independently validated; report composes their results.

Weighting

Composite score = score × 0.5 + drift × 0.3 + readiness × 0.2. Score is weighted highest because it is the broadest quality measure; drift is second because AI-generated UI drift is the 2026 crisis; readiness is third because machine-readiness is emergent and still maturing.

Shareability

Report is designed as the share hook — "here is your design-intelligence report" is more compelling than "run these six tools". One grade, one URL, one page.

URL gap

No competitor synthesizes audit + drift + readiness into a single composite report from a live URL — the designesy pattern.

Method

How it works

Fetch the target URL once. Fire /api/score, /api/drift, and /api/readiness in parallel (Promise.all). Each engine independently fetches the URL, extracts CSS + tokens, and runs its own checks. Collect the three results. Compute a composite score using the weighted formula: score × 0.5 + drift × 0.3 + readiness × 0.2. Derive a composite grade from the composite score. Aggregate all checks across the three engines into a single check list. Surface the three sub-scores, the composite score, the composite grade, and the full check inventory.

Weighting

01score × 0.540-check design-intelligence audit — the broadest quality measure (tokens, typography, motion, accessibility, anti-generic, runtime)

02drift × 0.312-check AI-generated UI drift radar — token fabrication, value variance, off-contract patterns

03readiness × 0.210-check AI readiness — machine-readable tokens, llms.txt, agent.json, MCP, DESIGN.md, sitemap

compositeScore = round(score × 0.5 + drift × 0.3 + readiness × 0.2). compositeGrade from compositeScore: A≥90, B≥80, C≥70, D≥60, F<60.

Verification

8 synthesis checks — 8 synthesis checks (rp01–rp08). The composite score is NOT derived from these checks — it is the weighted average of the three sub-engine scores. These checks verify the synthesis itself ran correctly. PASS=1, WARN=0.5, FAIL=0. Composite grade: A≥90, B≥80, C≥70, D≥60, F<60.

01rp01 · Target URL fetched and validatedPASS: URL is a public http(s) URL and was accepted by the guard · FAIL: URL is invalid, private, or blocked by the SSRF guard

02rp02 · Score engine completed — 40-check audit ranPASS: Score engine returned a result with a score and grade · FAIL: Score engine failed or timed out — the audit did not complete · WARN: Score engine returned but with SKIP checks — partial result

03rp03 · Drift engine completed — 12-check drift radar ranPASS: Drift engine returned a result with a score and grade · FAIL: Drift engine failed or timed out — the drift radar did not complete

04rp04 · Readiness engine completed — 10-check readiness probe ranPASS: Readiness engine returned a result with a score and grade · FAIL: Readiness engine failed or timed out — the readiness probe did not complete

05rp05 · Composite score computed — weighted synthesisPASS: Composite score computed from all three sub-scores using the weighting formula · FAIL: Composite score could not be computed — one or more sub-scores are missing · WARN: Composite score computed from partial results — one or more engines returned no score

06rp06 · Composite grade derivedPASS: Composite grade derived from the composite score (A–F) · FAIL: Composite grade could not be derived — composite score is missing

07rp07 · Check inventory aggregated — all checks across engines collectedPASS: All checks from score, drift, and readiness collected into a single inventory · FAIL: Check aggregation failed — engine results are malformed

08rp08 · Report is coherent — no engine contradicts the compositePASS: No engine score is more than 30 points from the composite — the grade is defensible · FAIL: One or more engine scores are more than 30 points from the composite — the grade is not defensible · WARN: One engine score is 20–30 points from the composite — the grade is borderline

Validation: Designesy report engine — parallel /score + /drift + /readiness with weighted synthesis. Method: Fire three internal API calls in parallel, collect results, compute composite score, derive composite grade, aggregate checks. Browser-only checks: None — all checks are server-side API composition, no browser needed.

Relationship to core

/score

Report runs /score as its primary sub-engine (weight 0.5) — the 40-check audit is the backbone of the composite

/drift

Report runs /drift as its second sub-engine (weight 0.3) — drift radar feeds the composite

/readiness

Report runs /readiness as its third sub-engine (weight 0.2) — readiness feeds the composite

/guardrails

Report does not run guardrails — guardrails emits a build contract, not a score; it is a companion, not a sub-engine

/monitor

Report is a single-point snapshot; /monitor tracks the snapshot over time

/compare

Report scores one URL; /compare diffs two — report is the input to compare

Open questions

The weighting (0.5/0.3/0.2) is a design decision, not a measured optimum — it may need tuning as the three engines mature at different rates
Report fires three internal API calls, each of which fetches the target URL independently — the target site sees three fetches per report, which may trigger rate limits on strict origins
The coherence check (rp08) flags divergence but does not resolve it — a site that scores A on tokens but F on drift gets a C composite, which may mislead
Report does not include guardrails or monitor — a future v0.2 could add a 4th weight for guardrail conformance
The composite grade is a single letter — it hides the sub-engine grades; the report UI must surface all three alongside the composite

Run it

Generate a unified design-intelligence report for any URL:

Generate a report →API endpoint

The report contract is the synthesis capstone — it answers “how good is this design, is AI breaking it, and can agents use it” in one composite grade. Machine export.