Primitive 0101
Core Loop
The smallest sequence of actions a user performs to receive the value you promised. Everything outside this loop is a candidate for the cut.
A skill for Claudev2.1 — now with memoryMIT · open source
When to building.
enough is a Claude skill that tells you the truth about your scope — and now it remembers what it cut. Feed it an idea, a half-built project, or a feature request. It returns a verdict, not a discussion.
Six modes. A persistent ledger. Drift audits that quote your prior cuts back when you try to reverse them. Domain packs for B2B SaaS, consumer, and creator products.
0
Modes of intervention
0
Cut-test patterns
0
Domain packs
0
Verdict, no debate
§ 01 — The problem
You don't have a scope problem. You have a stopping problem.
Every feature feels essential. Every edge case feels urgent. The codebase swells, the deadline slips, and the thing you set out to ship turns into the thing you're afraid to release.
And the cut you made in March? Forgotten by July. Re-litigated in August. Half-built by October. That's not scope creep — that's memory failure.
fig. 01scope, in its natural state
§ 02 — Primitives
Five rules. No exceptions.
The skill is opinionated because soft frameworks produce soft verdicts. These five primitives sit underneath every cut, ship, and kill it issues — and they are the same in every domain pack.
Primitive 0202
Loop Tax
Every feature taxes future builds, future bugs, and future decisions. If the tax doesn't widen the loop's reach, it's debt — not progress.
Primitive 0303
Evidence Ladder
Five rungs from "I think" to "users pay." No feature gets built off a single rung-1 opinion. The skill names which rung you're on.
Primitive 0404
Definition of Done
Written before the cut, signed under your name. A stranger uses the core loop without you in the room. That's the bar. Nothing else.
Primitive 0505
One Metric
One number. Reviewed weekly. Every triage answers "does this move the metric?" If the answer is no, the answer is no.
The mandate
"Enough is the smallest version that lets a stranger use the core loop and pay you — and nothing else gets built until that stranger exists."
— SKILL.md · L9
§ 03 — How it works
Three steps, one answer.
Same flow every time. No interview, no roadmap workshop, no "let's explore." A verdict, with the work to do underneath it.
Step 01
Describe
Tell the skill what you're building, where you are, and what you're tempted to add next. It reads any prior verdicts from your .enough/ ledger first.
Step 02
Detect
It runs your input through 18 cut-test patterns, the 5 primitives, and your domain pack — then names which rung of the evidence ladder you're actually on.
Step 03
Rule
You get a verdict — ship, cut, kill, or watch — signed under your name and written back to the ledger. Future sessions audit drift against it.
§ 04 — The output
A verdict.
Not a discussion.
Every session ends with a signed markdown file. Future you can read it. Future Claude reads it before it answers. The same question never costs you the same hour twice.
.enough/2026-06-12-ship.md● signed
verdict: SHIP after 9 hrs
blockers: 3 (payment, error msg, deletion)
felt blockers: 6 — none real
cuts: Beehiiv API, email reports, comparison view, chart colors, tests, import rework
DoD: a writer you've never spoken to uploads their export, sees their best send time, and pays — while you're asleep.
signed_by: tanuj · kill_if: <10 paying / 30 days
.enough/2026-09-04-drift-audit.md● audited
cuts checked: 6
cuts held: 4 · cuts breached: 2
· /src/integrations/beehiiv exists — breach. Promotion trigger (5 paying asks) never fired.
· /src/compare/ exists — breach. No paying user requested it.
verdict: REVERT both. Re-litigate only with rung-4 evidence.
quotes prior ship verdict · ledger-trail intact
§ 05 — Six modes
Same skill. Six places to plug it in.
The cost of building the wrong thing grows over time. enough dispatches to the mode that matches where you are — pre-commit, mid-build, post-launch, or back six months later to ask the harder questions.
Mode 01
Pre-Build
before commit
Pressure-test the idea against its smallest viable shape. Writes the first spec to the ledger and the kill criterion that ends it.
Mode 02
Mid-Build
halfway in
Halfway in and adding sub-features. The skill names what to cut today — and the trigger that would let you add it back.
Mode 03
Post-Launch
live, getting asks
Live, getting requests. It tells you which to ignore, which to do manually, and which are the product.
Mode 04
new · v2Drift Audit
did the cuts hold?
Returns to a prior project, reads the ledger, scans the repo, and quotes your own cuts back when you've quietly built them anyway.
Mode 05
new · v2Sunset
should this die?
Reads the kill criterion you signed. If the metric never came, the verdict is KILL — with a paragraph for the users and the postmortem.
Mode 06
new · v2Pricing
what to charge
Not a calculator. Names the value loop, the comparable anchor, and the smallest price that signals seriousness — then writes it to the ledger.
§ 06 — The Ledger
Cuts that remember themselves.
Every verdict is a markdown file in .enough/, signed under your name, with a deadline and a kill criterion attached. Future sessions read the ledger first — so the cut you made in March still holds the line in October.
your repo · .enough/● 7 files
.enough/├─ 2026-03-12-spec.md— Pre-Build spec · signed├─ 2026-03-12-ship.md— Ship Verdict · 9hrs├─ 2026-05-04-triage-mobile-app.md— Post-Launch · DO MANUALLY├─ 2026-05-04-triage-notion-export.md— Post-Launch · WATCH├─ 2026-06-21-pricing.md— Pricing Verdict · $9/mo├─ 2026-09-04-drift-audit.md— Drift Audit · 2 breaches└─ 2026-11-30-sunset.md— Sunset · KILL · postmortemspec.md · frontmatter● signed
---
id: 2026-03-12-spec
type: pre-build-spec
product: newsletter analytics
created: 2026-03-12
deadline: 2026-04-12
domain: b2b-saas
signed_by: tanuj
kill_if: <10 paying / 30d
---Mechanism 01
Sign on the way in
Every verdict carries your name, a date, a deadline, and the kill criterion. The line is set before the build starts.
Mechanism 02
Read on the way back
Future sessions open .enough/ first. Prior cuts ground the current answer. You can't re-litigate without naming what changed.
Mechanism 03
Audit, don't trust
Drift Audit grep-checks the repo for the cuts you signed. If you built it anyway, it quotes you back to yourself.
§ 07 — Domain packs
The cuts depend on the game you're playing.
What's enough for a B2B sales motion is not enough for a consumer habit loop, and neither matches a creator's audience-first build. Each pack rewires the cut tests, the evidence ladder, and the One Metric for its world.
Pack
B2B SaaS
for sellers, not builders
What it cuts
- ✕SSO before 10 customers
- ✕RBAC before 3 roles
- ✕SLA before $10k MRR
- ✕Custom auth
- ✕Dashboards before retention
What ships instead
One workspace. One role. One price. A weekly metric email.
Pack
Consumer
for habit, not novelty
What it cuts
- ✕Onboarding tours
- ✕Notifications
- ✕Streaks before retention
- ✕Social before utility
- ✕Profile pages
What ships instead
An empty state that's the product. A 60-second first value. One core loop.
Pack
Creator
for audience, not platform
What it cuts
- ✕Plugin marketplace
- ✕Collaboration
- ✕Themes
- ✕Custom domains before 100 readers
- ✕Analytics dashboard
What ships instead
Write button. Publish button. One subscribe link. A plaintext email digest.
§ 08 — Cut patterns
18 things you don't need.
The 18 patterns the skill checks against, with the manual version that ships this week. Each one carries a promotion trigger — the named condition that earns it back into the build.
01Admin dashboard
A spreadsheet
02User settings
Sensible defaults
03Onboarding flow
An empty state
04Notifications
An email
05Search
Cmd-F
06Filters
Two views
07Permissions
One role
08Custom auth
Magic link
09Realtime sync
Refresh button
10Drag & drop
Up/down arrows
11Dark mode
A light theme that's actually nice
12Mobile app
A responsive page
13Analytics dashboard
A weekly email to yourself
14AI feature
A really good form
15Workflow engine
An if-statement
16Multi-tenant
One workspace per account
17Plugin system
A copy-paste snippet
18Settings page
Delete the setting
§ 09 — Install
One clone. That's it.
Drop it into your .claude/skills directory or upload to Claude.ai. The ledger writes itself into your repo at .enough/. No accounts. No tracking. No upsell.
six
modes
one
verdict
zero
accounts
~/.claude/skills
$ cd ~/.claude/skills
$ git clone \
https://github.com/tanujrajputdev/enough
→ skill ready. ask claude: "use the enough skill on this."▍
Build less.
Ship sooner.
Remember why.
enough is free, open source, and one skill long. Take a verdict. Cut a feature. Sign the ledger. Go outside. The work is finished — and next month, it stays finished.