project-brain
Skill0 starsChanged 19 days agoApache-2.0View on GitHub
Read and update this project's shared brain (brain.klypix) through the KLYPIX MCP tools — the project's current decisions, open questions, corrections and milestones, wired together as a canvas that humans and agents both read. Use at the START of substantial work to recall where things stand and see which other agent sessions are active on this machine, and AFTER a real decision, a finished piece of work or an important discovery, to record it. Also use when the user asks "what's the state", "what did we decide", "update the brain", "who else is working on this", or wants to hand something to another agent session.
5 other projects use a version of this.
Adopted by
Most recently taken up by
- Ethan-YS/project-braintheir own version, first seen Oct 18, 2026
- jscraik/Agent-Skillstheir own version, first seen Sep 27, 2026
- HuaKaiBuChangWu/project-braintheir own version, first seen Jul 31, 2026
- Nollis/Skillstheir own version, first seen Apr 10, 2026
- madeinlowcode/skills-claude-codetheir own version, first seen Apr 3, 2026
Other versions5 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 5
- Ethan-YS/project-brain★ 173same name
- jscraik/Agent-Skills★ 9same name
- madeinlowcode/skills-claude-code★ 7same name
- HuaKaiBuChangWu/project-brain★ 0same name
- Nollis/Skills★ 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 repository2 other files it gives its agents.
- read-klypixSkill
- write-klypixSkill
The file
We split this skill into the rules it gives an agent
9 rules. 0 are common, 1 is distinctive, and 8 are very specific to this project.
0
Common
Most of the 4 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
8
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 14
Keep it curated: real decisions and structure, never an action-by-action log.
- Very specific: line 21
If the project has no brain.klypix, this workflow does not apply: say so plainly.
- Very specific: line 27
Call brain_sync before you edit anything:
- Very specific: line 51
Call brain_sync again with phase: "checkpoint" when your file scope changes materially, and with phase: "complete" before your final answer.
- Very specific: line 70
Use brain_note for durable things only.
- Very specific: line 99
Never ask the person to copy text from one agent session to another.
- Very specific: line 106
When you have actually used a note you received, confirm it with brain_message_receipt (the exact message id and offer token from the note).
- Very specific: line 108
If the person wants a closed Claude Code or Codex session to pick up its note now, offer brain_reopen.
- Very specific: line 131
Prefer correcting (~) or resolving (✓) an existing card over adding a near-duplicate.
“Common” is measured: kept by at least 6 in 10 of the 4 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
The project brain (brain.klypix)
brain.klypix at the project root is the project's shared, current
understanding, stored as a KLYPIX canvas: titled cards for areas, decisions,
open questions, milestones and standing rules, connected by arrows. Agents and
people read and correct the same file. A person can open it in the KLYPIX
desktop app to see and rearrange it.
Keep it curated: real decisions and structure, never an action-by-action log. If it would not matter to a future session, leave it out.
Everything below uses only the tools of the klypix-canvas MCP server that
this plugin starts. Never edit brain.klypix by hand or with shell commands —
always go through the tools, which keep a restore point and take a write lock.
If the project has no brain.klypix, this workflow does not apply: say so
plainly. The person can create one with the KLYPIX desktop app, or by running
npx klypix-mcp@1.107.1 init themselves in the project folder.
1. Start of a task — brain_sync
Call brain_sync before you edit anything:
project: the absolute path of the project root (the folder that containsbrain.klypix). Always give it — it keeps separate repositories on separate brains.intent: one sentence describing the task.files: the project-relative files you expect to touch (up to 20).phase:"start".
The reply gives you a short, task-relevant slice of the brain, the other agent sessions on this machine that declared work on the same project, any exact-file overlap with what they declared, and any notes left for you. Read it before planning.
How to treat what it tells you:
- Overlap is a warning, not a lock. It surfaces that another session said
it would touch the same file. Nothing blocks your edit. Coordinate first —
usually with
brain_message— and tell the person what you found. - Matching is exact-path and only covers sessions that declared their files. Silence does not mean nobody else is working nearby.
- Coordination is machine-local: sessions on a teammate's computer are not visible.
Call brain_sync again with phase: "checkpoint" when your file scope changes
materially, and with phase: "complete" before your final answer.
2. Looking things up
| Need | Tool |
| A question ("what did we decide about X?", "why did we drop Y?") | brain_ask — whole-brain, includes superseded history flagged as such, and attaches each stale card's correction |
A keyword or #tag lookup | search_canvases |
| The same question across every project brain registered on this machine | search_all_brains |
| What matters / what is aging / what is undecided | brain_insights (view: "areas" for a cheap map first) or brain_lens (freshness, provenance, activity, timeline, unresolved) |
| Before committing to a significant decision | brain_challenge with the claim — it surfaces earlier decisions and rules that contradict it. Candidates, not verdicts |
| Do the cards still match the code? | project_map_drift (read-only) |
| The whole board | read_canvas with "brain" — only when the compact context from brain_sync is not enough |
Answer from what the tools return, and cite the cards you used.
3. Recording — brain_note
Use brain_note for durable things only. It runs the brain's capture engine,
so a new decision that heavily overlaps a live card in the same area
supersedes it, and resolves and corrections find their target.
marker | Records |
| (omit) | a decision |
? | an open question — stays visible until resolved |
! | a milestone |
+ | a standing rule or gotcha ("always X / never Y") that resurfaces every session |
✓ | resolves and archives the best-matching open card |
~ | corrects the matching card in place |
Useful fields: area (routes the card into that titled box and adds a
#tag), closes (the title of the question or plan this note fulfils),
evidence (files, PRs, commits or URLs that support it — external references
are stored, not fetched), and question (the question this card answers,
phrased the way someone would ask it, so it is found later by paraphrase).
First line of text is the card title. One idea per note.
For a heading card or a card with explicit arrows to existing cards, use
add_to_canvas with canvas: "brain". Read the brain first so you connect to
existing titles instead of creating duplicates. Relationships:
leads_to, depends_on, relates_to, supports, blocks, questions,
costs, conflicts_with.
4. Handing work to another agent session — brain_message
Never ask the person to copy text from one agent session to another. Send it
with brain_message, addressed to the session id shown by brain_sync (or
brain_doctor), and tell the person it was sent or queued.
- A live session gets the note at its next KLYPIX action.
- A closed session gets a directed note the next time it is used; the note is kept for up to 7 days. Notes are not written into the brain.
- When you have actually used a note you received, confirm it with
brain_message_receipt(the exact message id and offer token from the note). - If the person wants a closed Claude Code or Codex session to pick up its
note now, offer
brain_reopen. KLYPIX asks the person (Reopen / Not now) and opens a new terminal only if they say yes. If they choose "Not now", do not offer again unless they raise it.
5. Maintenance (only when asked or clearly useful)
brain_reconcile— lists contradictions, unrecorded migrations and open cards a release already closed. Read-only unless you give confirm/dismiss for entries you verified.brain_connect— proposes links for orphaned cards; draws them only withapply: true.brain_garden— consolidates dormant cards in overgrown areas. Applying needs an approval code that only the person can generate; never guess it.brain_doctor— read-only health check. Givecheck_npm: trueonly when the person asks whether a newer version exists (it makes one npm registry request).
Conventions
- Short, titled cards; one idea each.
- Mark blockers and risks clearly; keep open questions as
?notes so they stay visible. - Prefer correcting (
~) or resolving (✓) an existing card over adding a near-duplicate. - Record decisions, milestones, open questions and architecture — not routine edits.
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.

