new-experiment
Skill0 starsChanged 20 days agoMITView on GitHub
Start a new experiment when the founder says "test the idea that X" - allocate the next number, create the folder, fill the test card with its structured header, then sync the dashboard (which regenerates the board).
2 other projects use a version of this.
Adopted by
Most recently taken up by
- Tresjs/trestheir own version, first seen Oct 8, 2026
- freemty/labmatetheir own version, first seen Sep 5, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- Tresjs/tres★ 3.8ksame name
- freemty/labmate★ 25same 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 repository6 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- get-startedSkill
- log-experimentSkill
- pull-evidenceSkill
- sync-dashboardSkill
- tourSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 1 is distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 6
Create one experiment for the idea the founder gave.
- Very specific: line 11
Who it is about: Read customer-segments.md and name the segment (Seg N) in the hypothesis.
- Very specific: line 12
If no real customer has been placed in a segment yet, write "nobody tagged to a segment" in the header's segment rather than guessing.
- Very specific: line 42
Canvas: If the box the test attacks is not in the canvas it belongs to (step 3), add a row for it there (raise version, set date, add a …
- Very specific: line 46
Sync: Run /sync-dashboard. The build fails with a PROBLEM: line if the header is wrong.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Create one experiment for the idea the founder gave. One hypothesis, one test, one metric. A hypothesis with "and" in it is two experiments: say so and ask which one first.
- Number and folder. The next ID is the highest
NNNN-folder underexperiments/+ 1, four digits. Createexperiments/NNNN-<slug>/(slug: 2-4 words, kebab-case). - Who it is about. Read
customer-segments.mdand name the segment (Seg N) in the hypothesis. If no real customer has been placed in a segment yet, write "nobody tagged to a segment" in the header'ssegmentrather than guessing. - Test card. Copy
experiments/_template/test-card.mdinto the folder astest-card.mdand fill both parts:- The header (the YAML block between the
---lines; the dashboard and the board are built from it):id,title(one line, no version notes),segment,theme,methodwritten exactly as#NN Namefromexperiments/METHODS.mdorCustom (closest: #NN Name),boxes(box IDs from either canvas:experiments/canvas/VP-A.mdfor the value side, one letter, J1, C4;experiments/canvas/LC-A.mdfor the business model, two letters, CH2, RS1; prefix~for the nearest box when there is no exact one),status: planned,decision: none,bar(the number that makes it a win),result: not started,effortandevidence(1-5),waiting_on(usually "Your go to start"),note(one line: where we are). - Which canvas. A job, pain, gain, product, pain reliever or gain creator is a
VP-Abox. Reach, retention, price, cost, a partner or supplier, a regulator, a key metric or an unfair advantage is anLC-Arow.themestays the card's own call: the Lean Canvas fixes no theme by block. - The card body: Hypothesis ("We believe that ...", segment named), Test, Metric, Criteria,
and why this is the cheapest believable test. Prefer cheap Discovery methods (#1 Customer
Interview, #12 Link Tracking, #13 Feature Stub, #14 404 Test) unless the idea is already
de-risked. If the test contacts people, add a "Before going live" checklist (per-person
tracking, test and internal accounts excluded, usable addresses, a pool not used by an earlier
experiment). Contact lists go in a CSV in the folder, never in the card: every
.mdfile is synced to the dashboard.
- The header (the YAML block between the
- How we measure. Fill the card's "How we measure" table before anything else is planned
around the test: for each source the test uses (the founder's own tools: an email service, a
database, web analytics, a form, interview notes), what it answers, the exact event, tag, filter
or query, whether the count is per person or overall, the baseline in a normal week, and the
dashboard dataset it becomes (
live-NNNNwhen it is a live query). If a connector for that tool is available in this session, check each source once now: the query runs, the events exist. Setsources:in the header to the sources the table uses. - Canvas. If the box the test attacks is not in the canvas it belongs to (step 3), add a row
for it there (raise
version, setdate, add a Changes line). A newLC-Arow also goes into one theme of that file's "Read by theme" table, or the build stops with aPROBLEM:line. Do not editPROGRESS.md: it is generated. - Sync. Run
/sync-dashboard. The build fails with aPROBLEM:line if the header is wrong. - Report to the founder in plain language: the hypothesis, the method and why it is the cheapest believable test, the bar, and what is their call before it starts (usually: the go).
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.

