lean
Skill0 starsChanged 46 days agoMITView on GitHub
Sizing a solution before code: the seven-rung ladder (YAGNI, codebase, stdlib, native, installed dep, one line, minimum code), the never-simplify list, debt: markers.
9 other projects use a version of this.
Adopted by
Most recently taken up by
- varelaia/claude-skills-publictheir own version, first seen Oct 26, 2026
- FareedKhan-dev/lean-agenttheir own version, first seen Oct 13, 2026
- mattbutlerengineering/skillstheir own version, first seen Oct 11, 2026
- utk2103/Prompt-Studiotheir own version, first seen Aug 22, 2026
- ousid/leantheir own version, first seen Jul 20, 2026
- naticorrea-03/claude-code-skillstheir own version, first seen Jul 1, 2026
Other versions9 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 9
- utk2103/Prompt-Studio★ 8same name
- naticorrea-03/claude-code-skills★ 4same name
- ousid/lean★ 4same name
- civillizard/claude-lean-skill★ 2same name
- CrystalPlatforms/ADM-CLI★ 1same name
- FareedKhan-dev/lean-agent★ 1same name
- mattbutlerengineering/skills★ 1same name
- Co4an/claude-auto★ 0same name
- varelaia/claude-skills-public★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository29 other files it gives its agents.
- hosts/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- adrSkill
- api-designSkill
- auditSkill
- code-reviewSkill
- concurrency-performanceSkill
- coverageSkill
- debuggingSkill
- debugSkill
- fast-laneSkill
- fixSkill
- implementSkill
- intakeSkill
- migration-safetySkill
- nonnaSkill
- observabilitySkill
- planSkill
- refactoringSkill
- releaseSkill
- reviewSkill
- rollbackSkill
- security-reviewSkill
- shipSkill
- supply-chainSkill
- syncSkill
- tddSkill
- tdd-workflowSkill
- testSkill
The file
We split this skill into the rules it gives an agent
4 rules. 0 are common, 4 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 8 edited versions we read keep these.
4
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 15
Stop at the first rung that holds.
- Distinctive: line 30
Only then: the minimum code that works — no speculative parameters, no config for a value that never changes, no layer for a future nobody …
- Distinctive: line 37
No interface with a single implementation, no factory for one product, no config knob for a value that never changes, no scaffolding "for …
- Distinctive: line 47
Not corners to cut, ever: validation at a trust boundary, error handling that prevents data loss, security, accessibility, and anything the …
“Common” is measured: kept by at least 6 in 10 of the 8 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Lean
Lazy means efficient, not careless — the goal is the smallest correct diff, never the least effort. The ladder below runs after you have read the task and the code it touches, not instead of reading it. Trace the real flow — who calls this, what already exists two files over, what the runtime already gives you for free — then climb.
The ladder
Stop at the first rung that holds; do not climb past it "just in case."
- YAGNI — does this need to exist? Name the lazier alternative in one line first. Half of "needed" code is speculative.
- Already in this codebase? Reuse it. Re-implementing something that lives a few files over is the single most common slop pattern — search before you write.
- The stdlib does it? Use it. Tested, documented, free; a hand-rolled equivalent is a maintenance liability with no upside.
- A native platform feature covers it? Use it —
<input type="date">over a date-picker dependency, a databaseUNIQUEconstraint over app-level checks. The platform ships more than most reach for. - An already-installed dependency solves it? Use it. Never add a new dependency for what a few lines, or something already in the lockfile, already does.
- Can it be one line? Make it one line — a comprehension or built-in beats a helper function that exists only to wrap it.
- Only then: the minimum code that works — no speculative parameters, no config for a value that never changes, no layer for a future nobody asked for.
Two rungs both hold? Take the higher one — it is closer to "does not exist."
Rules
- No interface with a single implementation, no factory for one product, no config knob for a value that never changes, no scaffolding "for later."
- Question a complex request in the same response, don't silently comply: "Did X; Y already covers the need. Want the full version? Say so."
- Two same-size stdlib options: take the one correct on the edge case, not the shorter one.
- The smallest change in the wrong place is still a second bug — lean is about size, not about dodging the real fix.
Never simplified away
Not corners to cut, ever: validation at a trust boundary, error handling that prevents data loss,
security, accessibility, and anything the requester explicitly asked for. Nonna's own gates are not
negotiable either: TDD with golden and property tests on the critical surface
(.claude/rules/testing.md), the machine-checked review verdict, and the five-mirror sync
(.claude/rules/sync.md). The ladder governs solution size, never the loop.
The debt: marker
Mark a deliberate shortcut so it stays visible instead of becoming permanent by accident:
debt: <ceiling>, <upgrade trigger> — a comma separates what you accepted from the condition that
should make someone revisit it.
# debt: global lock, per-account locks if throughput mattersBoth halves must be non-empty; a marker with no trigger after the comma is unfinished debt, not a
note. Gate: scripts/check-debt.sh. Default scans the tree and exits 1 if any marker lacks a
trigger. --range <git-range> scans only lines added in that diff — what /review runs, so a PR is
gated on debt it introduced, not on debt already in the tree. --ledger prints the grouped ledger
instead of gating — what /sync runs to check for drift. Usage error, or --range outside a git
repo, exits 2.
Review tags
Format: L<line>: <tag> <what>. <replacement>. Tags: delete: — adds nothing, remove it;
stdlib: — a standard-library call replaces a hand-rolled block; native: — a platform feature
replaces custom logic or a dependency; yagni: — built for a need that does not exist yet;
shrink: — same behavior, fewer lines.
L42: stdlib: 27-line email validator class. `"@" in email` (or the stdlib's own check).
L58: native: date-formatting library pulled in for one call. `Intl.DateTimeFormat`.
L71: yagni: `AbstractRepository` with exactly one implementation. Inline it.
L90: delete: retry wrapper around an idempotent local call that cannot fail transiently. Nothing.
L15: shrink: manual loop building a dict from two lists. `dict(zip(keys, values))`.Close with net: -<N> lines possible. when findings exist, or Lean already. Ship. when none do.
The tags apply to review asks too: a reviewer's speculative bound or i18n case with no failing input
is yagni: — answer it with a debt: marker, not code.
These findings carry category: simplicity and are MEDIUM at most — size never outranks
correctness or security.
You think you need X → the platform has Y
| You think you need | The platform has |
| Date picker library | <input type="date"> |
| Deep-clone library | structuredClone |
| UUID library | crypto.randomUUID() |
mkdirp | fs.mkdirSync(p, {recursive: true}) |
pytz | zoneinfo |
requests for one GET | urllib.request |
| App-level uniqueness check | a UNIQUE constraint |
| Pagination logic in code | LIMIT/OFFSET |
| JS animation library | a CSS transition |
| Modal dialog library | <dialog> |
lodash debounce | a 5-line closure |
moment for one format | Intl.DateTimeFormat |
| Hand-rolled deep-equal | the test framework's own assertion |
| Manual env-var loader | the platform's built-in env access |
Output
Lead with the code, not the reasoning. What you deliberately skipped, and the trigger to add it later, goes under Remaining risk in Nonna's four-line report — never its own heading.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

