north-star
Skill0 starsChanged 6 days agoMITView on GitHub
Distill a pile of wants into one north star sentence, with the Five W's beneath it.
13 other projects use a version of this.
Adopted by
Most recently taken up by
- uzysjung/uzys-agent-harnesstheir own version, first seen Nov 8, 2026
- lifekit-hq/devclawtheir own version, first seen Oct 21, 2026
- andreaskelm/pm-braintheir own version, first seen Oct 17, 2026
- fusebase-dev/fusebase-flowtheir own version, first seen Oct 16, 2026
- agentmart/pm-ostheir own version, first seen Oct 15, 2026
- poponline63/north-startheir own version, first seen Oct 13, 2026
Used with
Projects with it also use strategy.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
EpistatesSkillstrategy
Plan weekly or monthly X content strategy. Use when the user wants to plan their content calendar, set growth goals, build a content pipeline, or think strategically about their X presence.
Also in 3 projects that use this
Other versions13 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 13
- andreaskelm/pm-brain★ 235same name
- fusebase-dev/fusebase-flow★ 9same name
- ai-analyst-lab/north-star★ 8same name
- BhaskarManam/pstack★ 4same name
- raphaelDkhn/worldbuilder★ 3same name
- uzysjung/uzys-agent-harness★ 3same name
- ele-yufo/unstuck★ 2same name
- Naoray/skills★ 2same name
- poponline63/north-star★ 2same name
- agentmart/pm-os★ 1same name
3 more not listed
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 repository9 other files it gives its agents.
- AGENTS.mdAGENTS.md
- code-reviewSkill
- find-the-coreSkill
- glen-modeSkill
- show-meSkill
- start-with-whySkill
- ste-reportSkill
- wait-whatSkill
- wet-or-drySkill
The file
We split this skill into the rules it gives an agent
3 rules. 0 are common, 2 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 11 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 44
Read the pile back as a list.
- Distinctive: line 61
Two stars is no star: If the pile genuinely splits into two unrelated whys, say so plainly and split it: one north star each, and the user …
- Very specific: line 79
Read references/NORTH-STAR-FORMAT.md and draft the document.
“Common” is measured: kept by at least 6 in 10 of the 11 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
North Star
The user arrives with wants: things they want to build, fix, have, or become. Wants are surface. Under each want sits a need, the reason that want matters, and under the needs sits a single why. This skill ladders from the pile down to that why, states it as one sentence, and writes it down.
A compass, not a plan. The output points; it does not sequence. When how and when crowd in, note them on the side and carry on — the star has to outlive the plan.
Interview relentlessly. The user will offer the first plausible answer and want to be finished; that answer is almost always a want restated. Keep pushing.
Rounds
Work in rounds. Number each question and put your own recommended answer beneath it, so the user corrects rather than composes. One round per message: wait for the answers before the next round.
When a round is a genuine choice between options, ask it with a tappable-options tool if one is available — choosing costs the user less than typing. Keep the laddering rounds typed: there the exact wording is the data, and options would put your words in the user's mouth.
❓ **Q1** — **<title>**: <question>
➡️ <your recommended answer>1. Read before you ask
A question whose answer is already written down is worse than no question. Before the first round, exhaust what you can reach:
- An existing
NORTH-STAR.md,MISSION.md,CONTEXT.md, README, or planning notes in the workspace - The conversation so far, and anything the user attached
- The codebase, which often states what someone is actually building more honestly than a hurried reply will
Bring what you found into round one as a draft pile the user edits, rather than a blank page they fill. An existing star is a revision, not a fresh start: open by asking what moved.
Done when you have either a draft pile from the sources, or a specific reason there was nothing to find.
2. Empty the pile
Get every want out, unranked and unjudged. Ask what they want, then ask what else, until they are empty. Push for the wants that feel too small, too obvious, or too embarrassing to say — those carry the most signal.
Read the pile back as a list. Done when the user recognises the list and has nothing to add.
3. Ladder each want to bedrock
Take a want. Ask why it matters. Take that answer and ask why it matters. Repeat until the answer stops moving — the same reason coming back in different words. That is bedrock: a need.
Catch two things while laddering:
- A how in a why's clothing. "I need a knowledge graph" names a mechanism. Ask what the mechanism buys and it drops away, leaving the need.
- Someone else's why. Some reasons are inherited — an employer's, an industry's, a peer group's. Keep them, and mark them: the user should see which needs they hold and which they are carrying for someone else.
Done when every want in the pile traces to bedrock, and one more push on that bedrock returns the same answer twice.
4. Find the one why
Cluster the bedrock needs. Most of them turn out to serve a smaller number, and often one. Present the clustering and name the candidate why: the need the others are in service of.
Two stars is no star. If the pile genuinely splits into two unrelated whys, say so plainly and split it: one north star each, and the user picks which one this session is for.
Done when the user agrees on a single candidate why.
5. Test the candidate
Five tests. A fail sends you back to step 3 or step 4 — it does not get waived.
- Inversion — would anyone sane claim the opposite? If nobody would, it is a platitude ("do good work", "deliver value") and it steers nothing. Rewrite until the inverse is a position a real person holds.
- Forbid — name at least two wants from step 2 that this star says no to. A star that forbids nothing is decoration.
- Signature — could someone else in the user's position sign it unchanged? Then it is the field's why, not theirs. Sharpen until it is theirs.
- Live decision — take three decisions the user has open right now and run each through the star. It must change or confirm at least two. If it moves none, it is not operative.
- Survival — would it still hold if the current job, tool, project, or plan vanished? If not, it is a how. Ladder further.
Done when all five pass and you can show the user the evidence for each, test by test.
6. Write it
Read references/NORTH-STAR-FORMAT.md and draft the document.
Show the draft and hand control back before writing anything to disk: the star is the user's to sign, not yours to file. Ask whether it lands or what to change, and where it should go — defaulting to ./NORTH-STAR.md.
Once they sign it, write the file, then read the star sentence back on its own with none of the scaffolding. It has to stand up alone.
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.

