Compendium · Compositions · Score Rubrics

Score Rubrics

Items

of 1002 total

Full-featured composed scenes — one per row so each design has room to breathe. Click for the live render, the spec, and the source.

Featured 14 Atoms 762 Compositions 240

№ 1301

ICS Explainer

Scene STA do

DO · Rubric ICS · v2

Intent Completion Score
four dimensions. one number.

Aggregate

79/100

DEVELOPING

C Completeness

w · 30

All five canonical sections present and non-empty.

Sample sub-score 26/30

T Testability

w · 25

Each acceptance criterion is independently verifiable.

Sample sub-score 21/25

U Unambiguity

w · 25

No "should", "where appropriate", or pronoun ambiguity.

Sample sub-score 18/25

X Threat coverage

w · 20

STRIDE classes named, abuse cases enumerated.

Sample sub-score 14/20

// Formula


ICS = 1.0·C(30) + 1.0·T(25) + 1.0·U(25) + 1.0·X(20)

band(ICS): READY ≥ 85 · DEVELOPING ≥ 60 · UNFIT < 60

№ 1302

CCS Explainer

Scene STA go

GO · Rubric CCS · convention clarity

A convention is
a testable rule, not a slogan.

CCS

80/100

REVISE

Worked example "ALL files MUST use kebab-case"

scored against 4 dims ↓

P Precision

5/5

Words used have one defined meaning.

D Detectability

4/5

Violations are surfaceable by a linter, not a code review.

K Completeness

3/5

All cases covered — edge cases named, not punted.

S Scope clarity

4/5

Says exactly which files / domains it applies to.

№ 1303

SDS Explainer

Scene STA go

GO · Rubric SDS · STRIDE coverage

A threat without a mitigation
is a debt with a payment date.

SDS coverage

70%

PARTIAL

Asset	S	T	R	I	D	E
auth-service	●	●	◐	●	◐	●
spec-store	●	●	●	◐	◐	◐
agent-runtime	◐	●	—	●	●	●
audit-log	●	●	●	●	—	◐
webhook-bus	◐	◐	—	◐	●	—
column %	80	90	50	80	60	60

№ 1304

CSG Explainer

Scene STA go

GO · Guard CSG · constraint symmetry

Two specs. One impossible truth.

auth-session SPEC-019 · v2.1.0

## Constraints

- Token TTL: 8h hard ceiling

- Refresh window: 1h before expiry

- Issued by: auth-service only

≠ symmetric
contradiction

mobile-resume SPEC-022 · v1.4.0

## Constraints

- Offline session: 24h minimum

- Cached spec data: 7d

- Refresh on network return

Merge blocked

8h ≤ TTL ≤ 8h cannot coexist with 24h ≤ TTL — both surfaces resolve to the same token.

CSG · BLOCK

// Suggested resolution

Split the surface: introduce a separate offline-grant token, keep session at 8h. Then re-run CSG.

№ 1305

SHS Explainer

Scene STA do

SHS · Corpus health org · acme · 71 specs

refresh · 14:02:14 UTC

// Distribution

74/100

0–19

20–39

40–59

60–84

85–100

// Stale > 120 days

SPEC-007 184d @author UNFIT
SPEC-019 156d @author DEVELOPING
SPEC-031 142d @governor DEVELOPING
SPEC-044 121d @author READY

78 ▲ 2

71 ▼ 3

66 ▲ 1

82 ▲ 4

№ 1306

SCR Explainer

Scene STA go

GO · Radar SCR · supply chain

Dependencies have a pulse.
measure it before you trust it.

SCR aggregate

WATCH

// CVE coverage

scanned 84%

last scan · 41 min ago

// Vitality · release / maint / commit

astro 92·88·95
tailwindcss 89·76·84
react 81·90·78
left-pad-fork 12·8·14
crypto-shim 44·28·31

№ 1307

Two-stage Quality Gate

Scene STA go

GO · Pipeline two-stage gate

Structure first. Score second.

A high score on a malformed spec is a more confident wrong answer. Structure is the prerequisite, not a dimension.

Stage 1 · Structural binary

✓ frontmatter present & well-formed
✓ five canonical sections non-empty
✓ no dangling spec refs
✗ no TODO / TBD tokens

Stage 2 · Rubric numeric

ICS 86/100
CCS 79/100
SDS 71/100

Reject

no score awarded

a malformed spec
does not get
a low score —
it gets no score

№ 1308

Zero-Dimension Veto

Scene STA go

GO · Rule zero-dimension veto

High aggregate. Empty dimension.
Veto wins.

aggregate

74/100

VETO

C Completeness 28/30

T Testability 22/25

U Unambiguity 24/25

X Threat cov. 0/20

ZERO

Dimension `Threat cov.` scored zero.

Whole-rubric VETO triggers regardless of aggregate. A spec cannot ship without ANY threat coverage — security is not optional to make up later.

policy · GOV-002

№ 1309

Threat Floor

Scene STA go

GO · Rule threat coverage floor

Below the floor, nothing ships.

Per-dimension minimums force authors to address security at design time, not after launch.

blocked

2/5

FLOOR · 12

// Threat coverage · five specs

SPEC-014

18
SPEC-019

14
SPEC-021

12
SPEC-024

8
SPEC-029

4

floor at x = 12 / 20

№ 1310

Rubric Anti-pattern

Scene STA go

!! DO NOT rubric misuse · v1

Thresholds do not replace judgment.

01

Score-shopping

DON'TSpecs rewritten purely to inch the number above the band threshold without changing intent.

DOScore is a signal — not the artefact. Edit intent, not adjectives.
02

Threshold as escape hatch

DON'TA spec at exactly 60.0 is "DEVELOPING but shippable" — the band edge becomes the new normal.

DOBands are bins, not pass-marks. Edge cases trigger human review, not silent merge.
03

Human off the loop

DON'TAuto-merge any READY-band spec. No governor read-through, no risk note, no record.

DOScore gates filter — they do not approve. A human is on every irreversible step.

№ 1311

Configurable Thresholds

Scene STA go

GO · Config per-team thresholds

Same rubric. Different bindings.

Regulated work cannot calibrate the same way as prototype work. The numbers are the same. The thresholds are not.

pay-rails

regulated · external

team / 02

ICS READY ≥ 92 · DEV ≥ 75
CCS READY ≥ 88 · DEV ≥ 70
SDS READY ≥ 95 · DEV ≥ 80

lab-sandbox

prototype · internal

team / 02

ICS READY ≥ 70 · DEV ≥ 50
CCS READY ≥ 65 · DEV ≥ 45
SDS READY ≥ 60 · DEV ≥ 35

// Delta · pay-rails − lab-sandbox

ICS +22 READY · +25 DEV
CCS +23 READY · +25 DEV
SDS +35 READY · +45 DEV

№ 1312

Baseline Calibration

Scene STA go

GO · Onboarding baseline calibration

Three steps before a single threshold fires.

01 Sample

Pick 20 random specs from the last 6 months.

outcome
representative cross-section, not cherry-picked exemplars
02 Score

Score them all by hand. Two reviewers per spec, blind to each other.

outcome
inter-rater drift becomes visible, dimensions tightened
03 Recalibrate

Set the team's READY threshold at the 75th percentile of observed scores.

outcome
stretches but does not break — top-quartile becomes the floor

// Before · uncalibrated

75th pct ≈ 71 · noisy

// After · calibrated

75th pct ≈ 82 · sharpened

№ 1313

Risk-Stratified Review

Scene STA go

GO · Matrix risk × score = depth

Same score. Bigger blast. Deeper review.

Risk \ Score	UNFIT	DEVELOPING	READY
HIGH	HARD BLOCK	GOV + AUTHOR	DUAL SIGN-OFF
MED	HARD BLOCK	GOVERNOR READ	PEER REVIEW
LOW	REWRITE	PEER REVIEW	AUTO-MERGE

auto-merge

peer review

governor read

dual sign-off

hard block

№ 1314

Sampling Strategies

Scene STA do

DO · Practice sampling strategies

You will not score them all.
Pick a sampling strategy on purpose.

Systematic every 4th

// pick pattern · n=24

when:continuous health monitoring

Random stratified

// pick pattern · n=24

when:unbiased corpus snapshots

Spot-check governor-flagged

// pick pattern · n=24

when:incident response · drift

№ 1315

Score Trends

Scene STA do

DO · Trends 12 weeks · 6 rubrics

A score is a point.
A trend is the story.

period · 2026-03-01 → 2026-05-21

AAR ▲ 6

After-action recall 82

GTR ▼ 4

Green-to-red 74

SDL ▲ 5

Spec design lint 77

DDL ▲ 5

Drift detection 85

ICS ▲ 5

Intent completion 83

CCS ▲ 3

Convention clarity 70

№ 1316

Score-to-Action Map

Scene STA go

GO · Pipeline score → action

A number that triggers nothing
is just decoration.

Band · range Action Owner SLA d

Approved 85 – 100

A auto-merge to mainline

@gatekeeper 0

Ready 70 – 84

H governor sign-off + ship

@governor 1

Developing 40 – 69

H return to author with notes

@author 3

Unfit < 40

H rewrite from intent — do not patch

@author + @governor 7

Automated band. Only Approved is allowed to auto-merge. Everywhere else, a human owns the next move.

Human band. SLA starts when the band is entered. Missed SLAs escalate to the governor — not the author.

— Full detail →

—

// source

Alpha · Honest about it

The framework is real.
The community is forming now.

DoCoDeGo is in Alpha. The framework is documented, the practices are battle-tested at small scale, and the next release is being shaped in public.

If it produces anything, it should produce engineers and teams who think more clearly about what they are building and why.

Two doors in

Join the Discord

Discord is where specs are debated, the framework gets sharper, and decisions land in writing. The conversation is the artefact.

Score Rubrics

The framework is real.The community is forming now.

The framework is real.
The community is forming now.