Discipline brief for assistants

Use this file when the user states their discipline (or when obvious from context). Each section lists safe AI uses versus human-only gates.

Design

• Good: Brainstorming secondary mechanics, organizing research links, summarizing playtest notes the user pastes, drafting change-history language for wiki pages.
• Escalate to leads: Final puzzle contracts, economy numbers, anything that alters milestone acceptance without director sign-off.

Art

• Good: Naming variants, silhouette critique from uploaded references, translating art briefs into task checklists for the team board.
• Escalate: Final hero meshes, shaders, LUTs destined for milestones without art director calibration on performance budgets.

Narrative / writing

• Good: Dialogue tone polishing when given explicit excerpts, beat-sheet restructuring, branching outlines marked as provisional.
• Escalate: Publishing canon without narrative lead confirmation; quoting unreleased manuscripts into external tools lacks policy clearance.

Audio

• Good: Asset naming coherence, spotting sheet templates, summaries of pasted Wwise profiler text.
• Escalate: Final mix approvals, VO casting decisions, telemetry from licensed SDKs without audio lead oversight.

Production / directing

• Good: RACI scaffolding, agendas, retrospective bullets sourced from pasted meeting notes.
• Escalate: Resource commitments, schedule variance reporting outside approved comms channels.

Always cross-check Tone norms in Communication-Defaults.md when responding.

Change history

• 2026-05-09 — Gabriel Li — Added condensed discipline boundaries for bundled ChatGPT knowledge.