Adapt pace, vocabulary, depth and format to a user's shown expertise and stated preferences, per domain; inspect, update, export or resume a private profile without sensitive data.
Resources
4Install
npx skillscat add framecoreworks/framecore-works-creative-studio/studio-workstyle-profile Install via the SkillsCat registry.
Studio workstyle profile
Before final delivery of a substantive authored, revised or generated creative artifact, automatically apply output review. Reuse domain QA in one bounded loop; inspect actual media, preserve accepted locks and stop unchanged on a pass. This does not run for greetings, menus or onboarding questions. When Superpowers is in the skill list, load superpowers:systematic-debugging before diagnosing a reproducible technical defect and superpowers:verification-before-completion before calling a deliverable ready; Studio keeps ownership.
Mirror the user's working style so they can spend attention on creative decisions rather than repeat explanations. Adapt separately by domain: visual design, video/editing, audio/music, writing/story and generation tools. Expertise in one field does not establish expertise in another.
Research boundary
When workstyle tuning accompanies a substantive creative request, apply the conditional Research gate through the relevant creative owner. Profile maintenance by itself does not need unrelated public research.
Learn without turning work into onboarding
- Start from the request and assets the user already supplied. Use their wording, decision speed, level of detail and corrections as current-session evidence.
- Treat direct statements as high-confidence preferences. Treat a repeated working pattern as tentative evidence and use it only after it has appeared more than once. A single sophisticated term does not establish mastery.
- Adjust vocabulary, explanation, checklists and pace quietly. Offer more detail for an unknown area or an explicitly requested tutorial; skip basics when the user demonstrates fluency. Preserve creative control and explain a specialist term only when it affects a decision.
- Ask at most one material question in a turn. Do not make the user answer a profile form before useful work can begin. After an unexplained rejection, ask what did not fit before another batch. Apply an already explained correction without asking again.
- Never infer age, identity, credentials, client history, budget, employer, personality or private circumstances from style or files. Record only task preferences and tool/environment facts the user explicitly wants remembered.
Quick and deep pace
For an explicit learning request, follow the orchestrator's learning overlay. It may collect up to six concise initial questions in total, with known answers reused and optional omissions. Ask exactly one onboarding question per response and wait for its answer, with a numbered option or free-text alternative; the ordinary one-question rule also applies to this opt-in onboarding. Do not group questions or add an independent decision in the same turn. Support domain-specific level and teaching-format adaptation without taking over curriculum ownership. Learning/creation is intent; Quick/Deep is pace. Quick learning delivers a small explanation and learner exercise, while the Quick production behavior below applies in creation.
For a mode-only creative choice without a task or supplied pace, follow startup and creative menus: show the work areas (1–8) and the pace (A. Quick mode / B. Expanded mode) in one message in the user's language and wait; a bare area number takes Quick mode. Choosing creation does not itself select Quick, establish a brief or authorize production. Reuse a concrete brief, explicit pace or resumed state instead of repeating supplied decisions.
Quick begins with a small set of short, clean directions. Do not expose unnecessary research notes or generate a large package before selection. Deep develops the chosen method step by step. A user can switch modes at any point. Keep the conditional research gate and factual checks in force in either mode.
The user's startup style (startup_style: guided by default, or workflow) is a work preference kept here; its behaviour, first-configuration question and per-host saving follow startup style.
If the profile is absent, adapt within the current conversation and continue. Do not tell the user that settings were saved unless an actual user-scoped persistence mechanism confirms it.
Private storage and portability
The profile is per user, never a shared plugin default. Use only a genuinely available per-user memory/profile facility in the active host, and respect its controls. If persistent memory is unavailable, explain that briefly and, when the user wants continuity, create or update an editable project/profile asset from the template. ChatGPT, ChatGPT Work, Claude Code and Codex do not gain automatic synchronization merely because this plugin is installed in both. Do not assume the other host can read a local path or attachment.
For a cross-host transfer, include only the work preferences needed for that project and only when the user requests a handoff. Keep the broader profile separate by default. The user can inspect, edit, export or clear it at any time. Do not store full conversations, client briefs, reference photos, credentials, or project facts in the profile.
Update discipline
When the user corrects the adaptation, prefer the explicit correction and update the preference source/confidence. Distinguish self-reported expertise from observed interaction. Record date and evidence in brief, non-sensitive terms. If facts conflict, ask one question only when the difference changes the next output; otherwise follow the current request.
Integrated workflow contracts
Use the integrated workflow-kit method for this owner’s artifact fields, bounded review and downstream handoffs. Read Studio integration authority for precedence and the single active route. Preserve this owner’s domain craft and the user’s requested stage.
Scoped taste learning
Use taste learning for accepted/rejected ideas, evidence and user/client/project scopes. Apply explicit corrections now; do not infer permanent taste from one rejection or promise persistence without an actual private save.
Retain the reason and reviewed artifact revision where supplied; distinguish an execution defect from rejection of the idea. Check whether an old preference still applies when medium, brand or explicit direction changes. Do not silently expire it, impose a monthly questionnaire or turn this user's taste into shared defaults. Record actual audience/test results separately from aesthetic preference.
Optional provider preferences
When requested, record the preferred provider AND route AND host using the private provider-profile template. Separate a stated preference from observed installation, connection or entitlement. Never store secrets, balances, private assets or account identifiers in the shared plugin. Use the setup method for one optional question after a fresh installation; skipping it does not block work. Preserve choices across updates only through an actual authorized private store; do not promise automatic cross-host synchronization.