product-studio
Skill1 starsChanged 52 days agoMITView on GitHub
Run a persistent product-and-design operating system for multi-session website, app, brand, and digital-product work. Use when ChatGPT should maintain durable design state across sessions or agents, organize research and decisions, coordinate Design Maestro and specialist skills, create or update a .design workspace, manage design territories and human decisions, run readiness/quality gates, and produce implementation handoffs without losing product rationale. JEV/TypeSafe AI and Firecrawl are optional accelerators, never hard dependencies.
- Reads credentials
In this repository32 other files it gives its agents.
- grok-firstSkill
- opencode-firstSkill
- apple-design-accessibilitySkill
- apple-design-aiSkill
- apple-design-commerceSkill
- apple-design-connected-technologiesSkill
- apple-design-contentSkill
- apple-design-controlsSkill
- apple-design-foundationsSkill
- apple-design-gamesSkill
- apple-design-healthSkill
- apple-design-inputsSkill
- apple-design-iosSkill
- apple-design-mediaSkill
- apple-design-menusSkill
- apple-design-navigationSkill
- apple-design-patternsSkill
- apple-design-platformsSkill
- apple-design-presentationSkill
- apple-design-principlesSkill
- apple-design-privacy-identitySkill
- apple-design-spatialSkill
- apple-design-system-experiencesSkill
- design-maestroSkill
- forjaSkill
- temperaSkill
- bambu-h2c-slice-printSkill
- bambu-print-troubleshootSkill
- blender-print-readySkill
- fdm-print-preflightSkill
- ads-videoSkill
- tutorial-videoSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
43 rules. About 0 are common, 29 are distinctive, and 14 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈29
Distinctive · estimate
This author's own choices. Worth a look.
14
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 14
Use Product Studio for persistence and lifecycle management.
- Distinctive: line 15
Use Design Maestro for creative orchestration and design judgment.
- Distinctive: line 16
Use specialist skills for domain depth.
- Distinctive: line 25
If Design Maestro is unavailable, continue with the Product Studio workflow and apply the same product-first principles, but do not claim …
- Very specific: line 29
Treat .design/ as the durable source of truth for product/design decisions when a working directory exists.
- Very specific: line 30
Never replace approved decisions silently.
- Distinctive: line 31
Do not dump chat transcripts into the workspace.
- Distinctive: line 34
Keep product strategy, brand, UX, content, and implementation constraints connected.
- Very specific: line 39
Read existing .design/ state before proposing changes.
- Distinctive: line 40
Prefer small, explicit updates to the workspace after each meaningful phase.
- Distinctive: line 44
When a working directory is available, use this project-local structure:
- Very specific: line 63
Use scripts/init_design_workspace.py to create missing files without overwriting existing work.
- Very specific: line 65
Read references/design-state.md for ownership and update rules.
- Very specific: line 71
Read BRIEF.md, PRODUCT.md, DESIGN.md, DECISIONS.md, and QA.md first.
- Distinctive: line 72
Read other files only as needed for the current task.
- Distinctive: line 80
Create the smallest useful product frame.
- Very specific: line 85
Run in conversational mode.
- Very specific: line 101
Update BRIEF.md only when the project frame materially changes.
- Distinctive: line 116
Do not jump to visual styling when product logic is unresolved.
- Distinctive: line 120
Use research only when it can change a product/design decision.
- Distinctive: line 131
Use normal web/search first.
- Distinctive: line 133
If JEV is available, it may rerank/classify research fragments before synthesis.
- Distinctive: line 152
Use Design Maestro for creative exploration when available.
- Distinctive: line 177
Do not treat palette swaps as separate territories.
- Very specific: line 183
When a choice is made, append to DECISIONS.md:
- Distinctive: line 194
Do not hide a creative choice behind a JEV score.
- Distinctive: line 226
Document intent and roles, not only raw pixels.
- Distinctive: line 243
Use specialist skills for interaction, adaptation, Apple-platform guidance, or accessibility as needed.
- Distinctive: line 285
After implementation feedback, user testing, review, or a strategic change:
- Distinctive: line 287
update the source document that owns the decision
- Very specific: line 290
update HANDOFF.md if implementation expectations changed
- Very specific: line 295
Read references/integrations.md before use.
- Very specific: line 313
If TYPESAFE_API_KEY or the integration is unavailable, fall back to ordinary reasoning.
- Distinctive: line 317
Use Firecrawl only when the research problem benefits from deeper web extraction/crawling.
- Distinctive: line 327
If Firecrawl is unavailable, use the environment's normal web/search tools and continue.
- Distinctive: line 331
When Design Maestro is installed, invoke it for:
- Very specific: line 340
After the Maestro phase finishes, Product Studio must translate stable results into the appropriate .design/ documents.
- Distinctive: line 346
When sources disagree, use this order unless the user explicitly overrides it:
- Distinctive: line 355
Do not silently let an old territory override a newer approved design direction.
- Distinctive: line 365
avoid multiple agents editing the same source-of-truth file concurrently when possible
- Very specific: line 367
Use HANDOFF.md as the implementation-facing contract and DECISIONS.md as the durable history of major choices.
- Distinctive: line 380
Do not output this bookkeeping when it adds no value to a tiny task.
- Distinctive: line 384
Read only what the current phase needs:
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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
Product Studio
Operate as the persistent product/design system around a project. Maintain durable state, coordinate Design Maestro for creative direction, record important decisions, and keep research, brand, UX, content, QA, and implementation handoff aligned over time.
Product Studio is not a visual style and not a code generator. It is the project's design memory and operating model.
Core relationship
Use Product Studio for persistence and lifecycle management. Use Design Maestro for creative orchestration and design judgment. Use specialist skills for domain depth.
Product Studio (state + process)
-> Design Maestro (creative direction)
-> specialist skills
-> Product Studio (record stable decisions)If Design Maestro is unavailable, continue with the Product Studio workflow and apply the same product-first principles, but do not claim Design Maestro ran.
Non-negotiable rules
- Treat
.design/as the durable source of truth for product/design decisions when a working directory exists. - Never replace approved decisions silently. Record meaningful changes in
DECISIONS.md. - Do not dump chat transcripts into the workspace. Persist conclusions, evidence, rationale, constraints, and open questions.
- Separate facts, assumptions, experiments, and approved decisions.
- Preserve provenance for external research so future agents can verify it.
- Keep product strategy, brand, UX, content, and implementation constraints connected.
- Code is downstream of product/design decisions unless the user is explicitly working in an implementation phase.
- JEV and Firecrawl are optional accelerators. Their absence must never block the workflow.
- JEV can inform routing and readiness; it cannot become the creative authority.
- Human/user decisions control irreversible or identity-level choices.
- Read existing
.design/state before proposing changes. Never assume a blank slate when project state exists. - Prefer small, explicit updates to the workspace after each meaningful phase.
Workspace contract
When a working directory is available, use this project-local structure:
.design/
├── BRIEF.md
├── PRODUCT.md
├── RESEARCH.md
├── BRAND.md
├── DESIGN.md
├── EXPERIENCE.md
├── CONTENT.md
├── DECISIONS.md
├── QA.md
├── HANDOFF.md
├── territories/
├── references/
└── reviews/Use scripts/init_design_workspace.py to create missing files without overwriting existing work.
Read references/design-state.md for ownership and update rules.
Startup procedure
If .design/ already exists
- Read
BRIEF.md,PRODUCT.md,DESIGN.md,DECISIONS.md, andQA.mdfirst. - Read other files only as needed for the current task.
- Identify current phase, approved decisions, open questions, and stale assumptions.
- Continue from the existing state; do not restart discovery by default.
If .design/ does not exist
- Initialize the workspace.
- Seed
BRIEF.mdwith known facts and assumptions. - Create the smallest useful product frame.
- Continue into the appropriate workflow.
If no writable working directory exists
Run in conversational mode. Keep the same mental model and produce a proposed .design/ state as a handoff artifact if useful, but do not pretend files were persisted.
Lifecycle
Phase 0 - Orient
Resolve:
- project identity
- current artifact/state
- current phase
- target outcome
- constraints
- approved decisions
- unresolved decisions
Update BRIEF.md only when the project frame materially changes.
Phase 1 - Product frame
Maintain in PRODUCT.md:
- audience / user roles
- primary jobs
- business goal
- product promise
- primary success event
- value/proof model
- constraints and non-goals
- product principles
Do not jump to visual styling when product logic is unresolved.
Phase 2 - Research
Use research only when it can change a product/design decision.
Maintain RESEARCH.md as a synthesis, not a bookmark dump:
- question being investigated
- sources and dates
- findings
- confidence / limitations
- implications
- decisions influenced
Use normal web/search first. Escalate to Firecrawl for deeper crawling, multi-page extraction, or structured research sets when it adds value.
If JEV is available, it may rerank/classify research fragments before synthesis. Preserve citations/provenance.
Phase 3 - Product story and content
Maintain CONTENT.md with:
- core message hierarchy
- terminology
- primary claims
- proof requirements
- CTA language
- content modules
- voice constraints
- unresolved copy questions
Content structure should support product comprehension before SEO optimization.
Phase 4 - Design territories
Use Design Maestro for creative exploration when available.
Store each serious territory as a file under territories/, for example:
territories/
├── 01-precision-infrastructure.md
├── 02-editorial-intelligence.md
└── 03-autonomous-motion.mdEach territory should define:
- thesis
- product story
- composition / layout language
- typography character
- color logic
- imagery/illustration
- interaction/motion character
- signature moment
- differentiation
- tradeoffs / risks
- best-fit contexts
Do not treat palette swaps as separate territories.
Phase 5 - Decision checkpoint
For materially different territories or identity-level choices, keep the human/user in the decision loop.
When a choice is made, append to DECISIONS.md:
- date
- decision
- status
- rationale
- evidence
- alternatives considered
- consequences
- owner if relevant
Do not hide a creative choice behind a JEV score.
Phase 6 - Brand and visual system
Maintain BRAND.md and DESIGN.md.
BRAND.md owns:
- positioning-relevant personality
- visual metaphor
- identity principles
- logo/mark rules when applicable
- typography direction
- palette intent
- imagery direction
- voice/tone relationship
- forbidden/avoid patterns
DESIGN.md owns:
- grid/container logic
- hierarchy
- spacing rhythm
- semantic color roles
- typography roles
- surfaces/depth
- shape/radius rules
- iconography
- motion principles
- signature moments
- responsive design principles
Document intent and roles, not only raw pixels.
Phase 7 - Experience model
Maintain EXPERIENCE.md with:
- sitemap/screen map
- navigation model
- key flows
- state model
- primary/secondary actions
- feedback
- empty/loading/error/success states
- keyboard/touch/pointer behavior
- responsive transformations
- accessibility requirements
Use specialist skills for interaction, adaptation, Apple-platform guidance, or accessibility as needed.
Phase 8 - Reviews and gates
Store substantial review outputs in reviews/ and keep QA.md as the current summary.
Recommended gate families:
- Discovery ready: enough context to explore product/design directions.
- Direction ready: enough evidence and product clarity to choose a territory.
- Build ready: hierarchy, content, system, states, responsive behavior, and accessibility requirements are sufficiently specified.
- Ship ready: implementation has passed the relevant design/accessibility/content checks.
JEV may evaluate atomic gate criteria if available. Record the rubric and evidence. Do not use one vague "quality score".
Read references/gates.md.
Phase 9 - Handoff
Maintain HANDOFF.md as the current implementation contract.
It should include:
- goal and audience
- approved direction
- page/screen map
- content hierarchy
- design-system decisions
- interaction states
- responsive rules
- accessibility requirements
- SEO/AI-discovery requirements when relevant
- assets/references
- things not to change
- open questions
- implementation constraints
- recommended specialist skills for the next phase
Handoff is a decision document, not a transcript.
Phase 10 - Maintain
After implementation feedback, user testing, review, or a strategic change:
- update the source document that owns the decision
- append major changes to
DECISIONS.md - refresh
QA.md - update
HANDOFF.mdif implementation expectations changed - mark obsolete territory/review docs as superseded rather than deleting useful history without reason
Optional JEV layer
Read references/integrations.md before use.
Appropriate JEV roles:
- classify incoming intent
- suggest relevant specialist skills from the live catalog
- rerank research fragments
- classify review findings
- evaluate atomic readiness criteria
JEV must not:
- invent the product strategy
- choose the brand identity as an unquestioned authority
- generate the final design direction
- replace evidence or user research
- replace accessibility verification
If TYPESAFE_API_KEY or the integration is unavailable, fall back to ordinary reasoning.
Optional Firecrawl layer
Use Firecrawl only when the research problem benefits from deeper web extraction/crawling.
Examples:
- competitor homepage + product-page comparison
- multi-page content inventory
- structured category research
- extracting dynamic websites
- discovering a site's information architecture
If Firecrawl is unavailable, use the environment's normal web/search tools and continue.
Design Maestro integration
When Design Maestro is installed, invoke it for:
- product/design direction
- design territories
- specialist routing
- visual-system synthesis
- critique/review orchestration
- implementation handoff synthesis
After the Maestro phase finishes, Product Studio must translate stable results into the appropriate .design/ documents.
Product Studio owns persistence; Design Maestro owns creative orchestration.
State precedence
When sources disagree, use this order unless the user explicitly overrides it:
- latest explicit user decision
- latest approved entry in
DECISIONS.md - current owning document (
PRODUCT.md,BRAND.md,DESIGN.md, etc.) - verified external evidence
- specialist recommendation
- old exploration/review material
Do not silently let an old territory override a newer approved design direction.
Multi-agent behavior
For work distributed across agents:
- give each agent the smallest relevant
.design/subset - tell the agent which files are authoritative
- require returned findings to distinguish facts, proposals, and decisions
- merge stable outcomes back into the owning documents
- avoid multiple agents editing the same source-of-truth file concurrently when possible
Use HANDOFF.md as the implementation-facing contract and DECISIONS.md as the durable history of major choices.
Output behavior
For each substantial session, finish with a compact state summary:
- what changed
- which
.design/files were updated - decisions made
- decisions still open
- current gate/status
- recommended next phase
Do not output this bookkeeping when it adds no value to a tiny task.
References
Read only what the current phase needs:
references/operating-model.md- architecture and Product Studio vs Design Maestro.references/design-state.md-.design/file ownership and state rules.references/workflows.md- end-to-end workflows for websites, apps, redesigns, and reviews.references/gates.md- readiness and quality gate patterns.references/integrations.md- JEV/TypeSafe AI and Firecrawl optional integration rules.references/handoff.md- handoff contract and multi-agent patterns.references/examples.md- concrete usage examples.
Scripts
scripts/init_design_workspace.py- create missing.design/structure without overwriting existing state.scripts/check_design_workspace.py- report required files, empty files, and current workspace completeness.
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.

