brainstorm
Skill2 starsChanged 5 days agoMITView on GitHub
Work through an idea conversationally with Claude. All brainstorms are logged in the brain with any decisions, questions, insights, or tasks that emerge.
189 other projects use a version of this.
Adopted by
Most recently taken up by
- nshopik/dream-teamtheir own version, first seen Nov 16, 2026
- crysix/skillstheir own version, first seen Nov 15, 2026
- fullstackhouse/skillstheir own version, first seen Nov 15, 2026
- chuanglibs/Claude-Code-Game-Studios-Codextheir own version, first seen Nov 15, 2026
- yagizdo/quivertheir own version, first seen Nov 15, 2026
- automagik-si/genietheir own version, first seen Nov 15, 2026
Used with
Projects with it also use playtest-report, team-polish and 10 more.
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.
CoralGameSkillplaytest-report
Generates a structured playtest report template or analyzes existing playtest notes into a structured format. Use this to standardize playtest feedback collection and analysis.
Also in 9 projects that use this
CoralGameSkillteam-polish
Orchestrate the polish team: coordinates performance-analyst, technical-artist, sound-designer, and qa-tester to optimize, polish, and harden a feature or area for release quality.
Also in 11 projects that use this
lilangMaxSkillsoak-test
Generate a soak test protocol for extended play sessions. Defines what to observe, measure, and log during long play sessions to surface slow leaks, fatigue effects, and edge cases that only appear after sustained play. Primarily used in Polish and Release phases.
Also in 6 projects that use this
lilangMaxSkillteam-live-ops
Orchestrate the live-ops team for post-launch content planning: coordinates live-ops-designer, economy-designer, analytics-engineer, community-manager, writer, and narrative-director to design and plan a season, event, or live content update.
Also in 6 projects that use this
- +8 more from CoralGame
Other versions189 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 189
- Donchitos/Claude-Code-Game-Studios★ 39ksame name
- codeaholicguy/ai-devkit★ 1.7ksame name
- jtydhr88/ComfyTV★ 1.2ksame name
- event-driven-io/emmett★ 567same name
- zaxbysauce/opencode-swarm★ 532same name
- ZaxbyHub/opencode-swarm★ 531same name
- automagik-dev/genie★ 353same name
- automagik-si/genie★ 353same name
- namastexlabs/automagik-genie★ 353same name
- bkywksj/knowledge-base★ 346same name
179 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 repository24 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- 1passwordSkill
- brainSkill
- bridgeSkill
- claude-switcherSkill
- clear-cacheSkill
- commitSkill
- cross-artifactSkill
- ctoSkill
- diskSkill
- killSkill
- lookSkill
- parseSkill
- pbarSkill
- planSkill
- pruneSkill
- pushSkill
- reportSkill
- researchSkill
- swarmSkill
- vibe-auditSkill
- video-editorSkill
- webSkill
- wikiSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
15 rules. 0 are common, 13 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 157 edited versions we read keep these.
13
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 48
Before diving in, gather relevant context:
- Distinctive: line 50
Check the brain DB for related entries (decisions, questions, insights, prior brainstorms):
- Distinctive: line 56
Check the codebase if the topic relates to implementation.
- Distinctive: line 65
Ask questions: Don't just answer — probe.
- Distinctive: line 102
When the conversation reaches a natural stopping point (or the user signals done):
- Distinctive: line 104
Update the brainstorm entry's body with a summary:
- Very specific: line 120
Start a brainstorm session anchored to the current focus items.
- Distinctive: line 137
Create a brainstorm entry that references the focus items:
- Distinctive: line 159
Follow the same Steps 1, 2, 4, and 5 as New Brainstorm, with these modifications:
- Distinctive: line 170
Attack the assumptions, not the person: Every plan rests on beliefs about the market, the user, the tech, the timeline.
- Distinctive: line 173
Be specific about risk: Don't say "this could be risky." Say "if churn exceeds 30% in month 2, this unit economics model breaks because..." …
- Distinctive: line 174
Steel-man before you attack: Before poking a hole, show you understand why the idea is compelling.
- Distinctive: line 175
Every criticism carries a suggestion: Never raise a risk without proposing at least one way to address it.
- Distinctive: line 193
Ask the user what aspect they want to revisit or continue.
- Very specific: line 230
If .claude/kit.json has a rules."brainstorm" entry, read it and apply it as an additional instruction for this skill.
“Common” is measured: kept by at least 6 in 10 of the 157 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
Recent Brainstorms
!sqlite3 -separator ' | ' cowork/brain/BRAIN.db "SELECT '#' || e.id, e.created_at, e.title, (SELECT count(*) FROM logs c WHERE c.parent_id = e.id) || ' outputs' FROM logs e WHERE e.type = 'note' AND e.tags LIKE '%brainstorm%' ORDER BY e.created_at DESC LIMIT 5;" 2>/dev/null || echo "No brainstorms yet"
Recent Decisions & Insights (for context)
!sqlite3 -separator ' | ' cowork/brain/BRAIN.db "SELECT '#' || id, type, title FROM logs WHERE type IN ('decision','insight') AND status = 'active' ORDER BY created_at DESC LIMIT 8;" 2>/dev/null || echo "None yet"
Brainstorm
Work through an idea conversationally. Every brainstorm is logged in the brain DB as a note with brainstorm tag, and any outputs (decisions, questions, insights, tasks) are linked as children via parent_id.
Input
Optional argument: $ARGUMENTS
Dispatch:
focus-> start a Focus Brainstorm on the current focus items.hater <topic>-> start a Hater Brainstorm — constructive skepticism to pressure-test the topic.hater(no topic) -> ask the user what to pressure-test.- A topic or idea string -> start a New Brainstorm on that topic.
- A bare number (e.g.
42) -> Revisit brainstorm entry #42 and continue the conversation. list-> List all brainstorms with their output counts.- Empty -> ask the user what they'd like to brainstorm.
New Brainstorm
Step 1 — Create the brainstorm entry
Insert a note to anchor the session:
INSERT INTO logs (type, title, body, tags, importance)
VALUES ('note', '<topic title>', 'Brainstorm session started.', 'brainstorm', 5);Note the new entry's id — all outputs from this session link back to it via parent_id.
Step 2 — Set context
Before diving in, gather relevant context:
- Check the brain DB for related entries (decisions, questions, insights, prior brainstorms):
SELECT id, type, title, substr(body, 1, 200) FROM logs WHERE (title LIKE '%<keyword>%' OR body LIKE '%<keyword>%' OR tags LIKE '%<keyword>%') AND status = 'active' ORDER BY created_at DESC LIMIT 10; - Check the codebase if the topic relates to implementation.
- Briefly present what you found — prior thinking on this topic, related decisions, open questions.
Step 3 — Brainstorm conversationally
This is a back-and-forth, not a monologue. Follow these principles:
- Think out loud. Share reasoning, not just conclusions. Surface assumptions.
- Be opinionated but holdable. Offer a clear perspective, but hold it loosely. Push back on weak ideas; build on strong ones.
- Ask questions. Don't just answer — probe. "What if we..." and "Have you considered..." are more valuable than lengthy analysis.
- Stay concrete. Tie abstract ideas to specific implementation, user scenarios, or project constraints. Reference the codebase and brain entries when relevant.
- Diverge then converge. Early in the brainstorm, explore widely. As clarity emerges, narrow toward decisions and actions.
- Keep it moving. Each response should advance the conversation. If you're restating the same ground, pivot or propose a decision point.
Step 4 — Capture outputs as they emerge
As the conversation produces clear outputs, log them immediately — don't wait until the end. Each output is a new entry with parent_id pointing to the brainstorm note:
Decisions:
INSERT INTO logs (type, title, body, tags, parent_id)
VALUES ('decision', '<what was decided>', '<rationale>', '<tags>', <brainstorm_id>);Questions surfaced:
INSERT INTO logs (type, title, body, tags, parent_id)
VALUES ('question', '<the question>', '<context for why it matters>', '<tags>', <brainstorm_id>);Insights:
INSERT INTO logs (type, title, body, tags, parent_id)
VALUES ('insight', '<the insight>', '<supporting reasoning>', '<tags>', <brainstorm_id>);Tasks identified:
INSERT INTO logs (type, title, body, tags, pillar, parent_id)
VALUES ('task', '<actionable task>', '<context and acceptance criteria>', '<tags>', '<pillar>', <brainstorm_id>);Announce each capture briefly: "Logged decision #N: ..." so the user knows it's recorded without breaking flow.
Step 5 — Close the brainstorm
When the conversation reaches a natural stopping point (or the user signals done):
- Update the brainstorm entry's body with a summary:
UPDATE logs SET body = '<summary of what was discussed and concluded>' WHERE id = <brainstorm_id>; - Count and report outputs:
SELECT type, count(*) FROM logs WHERE parent_id = <brainstorm_id> GROUP BY type; - Present a brief wrap-up: what was explored, what was decided, what questions remain open, and what tasks were created.
- Recalculate tiers (
/brain, Recalculate tiers).
Focus Brainstorm
Start a brainstorm session anchored to the current focus items. Usage: /brainstorm focus
Procedure
- Load focus items:
SELECT id, type, title, body, tags, pillar FROM logs WHERE focus = 1 AND status = 'active'; - Load related context — recent decisions, insights, and open questions connected to these focus items:
SELECT id, type, title, substr(body, 1, 200) FROM logs WHERE status = 'active' AND type IN ('decision', 'insight', 'question') ORDER BY created_at DESC LIMIT 15; - Create a brainstorm entry that references the focus items:
INSERT INTO logs (type, title, body, tags, importance) VALUES ('note', 'Focus brainstorm: <combined topic from focus items>', 'Brainstorm session on current focus items: <list titles and IDs>.', 'brainstorm,focus', 7); - Present the focus items, their descriptions, and related context. Then ask the user which focus item to start with, or whether to tackle them together (if they're related).
- Proceed with brainstorming (Steps 3-5 from New Brainstorm above). Keep the conversation grounded in the focus items — when outputs emerge, note which focus task they advance.
Hater Brainstorm
Pressure-test an idea like a skeptical investor or CTO would. Usage: /brainstorm hater <topic>
The goal is not to kill ideas — it's to make them stronger by surfacing risks, edge cases, and gaps in thinking before they become expensive surprises.
Procedure
Follow the same Steps 1, 2, 4, and 5 as New Brainstorm, with these modifications:
Step 1 — Create the brainstorm entry with tags brainstorm,hater:
INSERT INTO logs (type, title, body, tags, importance)
VALUES ('note', 'Hater: <topic title>', 'Hater brainstorm — pressure-testing assumptions.', 'brainstorm,hater', 6);Step 3 — Brainstorm with a skeptic's lens. Replace the standard principles with:
- Default to doubt. Assume the idea has a fatal flaw you haven't found yet. Your job is to find it — or confirm it's not there.
- Attack the assumptions, not the person. Every plan rests on beliefs about the market, the user, the tech, the timeline. Name those beliefs explicitly, then stress-test each one. "This assumes X — what if X is wrong?"
- Ask the investor questions. "What's the failure mode?", "Who else has tried this and what happened?", "What's the second-order effect?", "What does this look like at 10x scale?", "Where's the hidden cost?"
- Find the edge cases. The happy path is easy. Probe the unhappy paths — the user who doesn't fit the model, the technical constraint that only shows up at scale, the regulatory angle nobody mentioned.
- Be specific about risk. Don't say "this could be risky." Say "if churn exceeds 30% in month 2, this unit economics model breaks because..." Concrete scenarios, not vibes.
- Steel-man before you attack. Before poking a hole, show you understand why the idea is compelling. Then explain why that's not enough.
- Every criticism carries a suggestion. Never raise a risk without proposing at least one way to address it. "This breaks if X" is incomplete — "This breaks if X; you could mitigate by Y or validate with Z" is the standard. Unproductive negativity is a bug.
- Flag research gaps honestly. When a risk or decision hinges on something neither of us actually knows — market data, technical benchmarks, competitor behavior, legal constraints — say so and suggest a
/researchtopic. Don't pretend to have the answer and don't assume the user does either. Before suggesting research, checkcowork/research/for existing reports that already cover the topic. If a report exists, reference it instead of proposing redundant work. - Converge on what survives. The output isn't a graveyard of killed ideas — it's a ranked list of risks with severity and mitigation options, a set of questions that need answers before committing, and the core thesis restated with its actual (not imagined) strength.
Revisit
Continue a prior brainstorm. Usage: /brainstorm <id>
Procedure
- Load the brainstorm entry and all its children:
SELECT id, type, title, body, tags FROM logs WHERE id = <id>; SELECT id, type, title, body, tags, status FROM logs WHERE parent_id = <id> ORDER BY created_at; - Present the state: original topic, summary, and all outputs (decisions, questions, insights, tasks) with their current status.
- Ask the user what aspect they want to revisit or continue.
- Resume brainstorming — new outputs still link to the same parent via
parent_id.
List
Show all brainstorms with output counts. Usage: /brainstorm list
Procedure
- Query:
SELECT e.id, e.created_at, e.title, (SELECT count(*) FROM logs c WHERE c.parent_id = e.id) as outputs, (SELECT group_concat(DISTINCT c.type) FROM logs c WHERE c.parent_id = e.id) as output_types FROM logs e WHERE e.type = 'note' AND e.tags LIKE '%brainstorm%' ORDER BY e.created_at DESC; - Present as a list with id, date, title, and output summary. Highlight any with open questions.
Global rules
- No permission needed — Execute immediately without asking.
- Log as you go — Capture outputs during the conversation, not after. The brain DB is the record.
- Link everything — Every output entry must have
parent_idset to the brainstorm entry's id. - Normalize tags — Split by comma, trim, lowercase, sort alphabetically, deduplicate, rejoin. Check existing tags (shown in dynamic context above) before inventing new ones.
- DB is truth — The DB is the only copy of the brain. Always query it.
- Decisions are permanent — Don't update a decision entry. If thinking changes, create a new decision that supersedes it (set
supersedescolumn, mark old onesuperseded).
Project overrides
If .claude/kit.json has a rules."brainstorm" entry, read it and apply it as an additional
instruction for this skill. Absent file or key means no overrides — that is the normal case.
jq -r '.rules."brainstorm" // empty' .claude/kit.json 2>/dev/nullDoes 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.

