Brainstorm
Codex Runtime
This skill is fully contained in the installed .agents directory. Resolve all linked resources relative to this SKILL.md; do not read workflow resources outside this packaged directory.
Before acting, read GAT Workflow. Read Directory Structure or Workflow Catalog only when their structure or ownership details are needed.
Use these packaged role profiles for every role-specific pass:
When the workflow says to spawn one of these roles, create a Codex subagent with that exact task name when collaboration is available. Give it the complete role profile, relevant source documents, confirmed decisions, template, and exact target paths; wait for it and review its result before continuing. This skill explicitly authorizes those named workflow subagents. If subagents are unavailable, perform the same pass yourself using the packaged role profile.
Treat the directory containing .agents as the repository root. All GAT project documents live under that root's gat/ directory. Use Codex filesystem tools and the environment's supported patch editor while preserving unrelated user changes.
Interpret AskUserQuestion in the procedure as a single concise user question. Use a structured choice UI only for a genuine 2-3 option decision; otherwise ask in ordinary dialogue and wait. Command examples use Codex skill syntax ($gat-*).
This skill explores a game concept through open-ended designer interview.
Spawn gat-designer for design reasoning and gat-artist for visual direction.
Pick questions from the angle table below — follow the conversation, not a script.
Phase 1: Resolve Mode
- If argument is
discuss→ Mode:discuss(no files written, exploration only) - If argument is a hint or empty → Mode:
design(produces foundational design docs and global art direction)
Check whether gat/overview/game.md, gat/overview/systems-index.md, or
gat/overview/art-direction.md already exist. If so, note them — the interview may
refine or replace existing decisions instead of starting from scratch.
Phase 2: The Interview
Core Rules
- One question at a time. Never batch. Wait for the answer before the next question.
- Provide a recommended answer with each question. Explain the design reasoning.
- Prefer open-ended questions. Let the user type free-form responses. Reserve
AskUserQuestion(multiple-choice) for when you are presenting concrete solution options and need the user to pick one. Most of the interview should be dialogue, not forms. - Pick angles, don't follow steps. Use the angle table as a menu. Jump to whatever dimension is most useful next — chase what's interesting or ambiguous.
- If a question can be answered by reading existing design files, read them instead of asking.
- Spawn
gat-designerwhen you need a design reasoning pass: drafting a core loop, proposing systems, evaluating a trade-off, or checking consistency. - Spawn
gat-artistwhen visual identity needs synthesis: art references, palette, readability, asset groups, production standards, or conflicts between style and gameplay clarity.
Drilling into Vague Ideas
When the user has a fuzzy idea — about the whole game or a single system — your job is to make it concrete through relentless, curious questioning. This is the core of the interview.
How to drill:
- When the user says something vague ("combat should feel impactful"), ask what specifically makes it impactful — is it animation, sound, damage numbers, controller rumble, enemy reaction, time-to-kill? Keep asking until the abstraction bottoms out in concrete mechanics.
- When the user proposes a system, ask about its boundaries. What does it NOT do? What system owns the adjacent responsibility? A system without edges is still fuzzy.
- When the user describes a player experience ("I want the player to feel lost"), ask what the game does to create that feeling. What does the player see, hear, and do? What information is withheld? What mechanics produce the emotion?
- When the user references another game ("like Dark Souls but..."), isolate what exactly they want to keep and what they want to change. The reference is a shortcut — unpack it.
- Ask about edge cases and failure states. What happens when the player ignores the system? What happens when they optimize it to the extreme? The answers reveal whether the system is understood or still hazy.
- Ask about the player's moment-to-moment decisions. If the user can't describe what choices the player makes inside the system, the system isn't clear yet.
- If an answer opens three new questions, pick the most foundational one first. Resolve dependencies before details.
Signals that something is still vague and needs more drilling:
- The user uses abstract adjectives without mechanics behind them ("fun", "smooth", "deep", "cool")
- A system is named but its inputs, outputs, and rules are undefined
- Two systems have overlapping or unclear boundaries
- The user can describe what the system IS but not what the player DOES in it
- Numbers are absent where they matter (how many? how long? how much?)
Seed Extraction
If a concept hint was provided, first spawn gat-designer to extract what the
hint already answers. Briefly summarize what's established so the user can
confirm or correct before diving in. Skip if no hint.
Interview Angles
Pick questions from any angle below. There is no fixed order — follow the thread that matters most at each moment. The table is a palette, not a checklist.
| # | Angle | Purpose | Example prompts |
|---|---|---|---|
| 1 | Genre & Style | Establish the game's design identity and reference points | What genre(s) does this live in? What games should it feel like mechanically? Real-time or turn-based? 2D or 3D? Single-player, co-op, or competitive? |
| 2 | Visual Direction | Establish the global art identity that will become art-direction.md | What should the game look like at a glance? Which art references fit or should be avoided? What palette, shape language, camera, readability, and production constraints matter? |
| 3 | Core Player Verb | Pin down the primary action the player repeats | What does the player actually DO moment-to-moment — shoot, build, explore, talk, craft, steer, command? What makes that action satisfying? |
| 4 | Target Feeling | Define the emotional experience | What should the player feel during play — tension, mastery, wonder, power, relaxation, social connection, fear, curiosity? When do they feel it most? |
| 5 | Fantasy & Role | Clarify who the player is in the world | What fantasy does the game fulfill? Who is the player — hero, commander, survivor, creator, investigator, merchant? |
| 6 | Scope & Constraints | Set boundaries early | Rough scope (jam, indie, commercial)? Platform? Timeline? Team size? Content rating? Hard constraints? Any visual production constraints like pixel art, low-poly, UI-heavy, asset reuse, or resolution limits? |
| 7 | Core Loop | Map the repeatable cycle that drives engagement | What's the 30-second loop? The 5-minute loop? The session loop? What pulls the player back in? |
| 8 | Systems & Mechanics | Explore what systems the game needs | What systems does the core loop imply? Which are essential vs. nice-to-have? What does each system depend on? |
| 9 | Progression & Goals | Define how the player grows and what they strive for | Short-term goals? Long-term goals? Skill tree or gear-based? Linear or branching? How does difficulty ramp? |
| 10 | Economy & Resources | Map currencies, sinks, and sources | What resources does the player manage? How are they earned and spent? Is there inflation risk? |
| 11 | Risk & Reward | Balance tension against payoff | What does the player risk losing? What do they gain for taking risks? Is failure interesting or just punishing? |
| 12 | Player Agency | How much control and choice the player has | Where do players make meaningful choices? Are choices tactical (moment-to-moment) or strategic (long-term)? Emergent or scripted? |
| 13 | Feedback & Juice | How the game communicates back to the player | How does the player know they did something right? What visual/audio hooks sell the actions? Screen shake, particles, sound? Which of those hooks should drive the global art direction? |
| 14 | Onboarding & Clarity | How the player learns the game | Tutorial or discovery? How do you teach without lecturing? What's the first thing a new player does? What must be readable instantly in the UI or scene? |
| 15 | Narrative & World | Story, setting, and tone | Is there a story? Player-driven or authored? What's the tone? How does the world reinforce the mechanics and visual identity? |
| 16 | Multiplayer & Social | Other humans in the experience | Cooperative, competitive, or solo with social features? Synchronous or asynchronous? How do players interact? |
| 17 | Replayability & Depth | What keeps players coming back | Procedural generation, build variety, difficulty modes, secrets? What's different on run #2 vs. run #50? |
| 18 | Accessibility | Who can play and how | Difficulty options? Color independence? Remappable controls? Reaction-time accommodations? What visual signals must not rely on color alone? |
| 19 | Monetization | Business model (if applicable) | Premium, F2P, subscription? If F2P, what's sold and does it affect gameplay? Any dark patterns to avoid? |
| 20 | Platform & Controls | Input method and platform constraints | Mouse/keyboard, controller, touch? How many buttons does the design assume? Platform-specific constraints? |
| 21 | Content Volume | How much stuff the game needs | How many levels, enemies, items, abilities? Is content hand-crafted, procedural, or both? What's the MVP slice? Which asset groups must be planned globally? |
Navigating the Interview
- Start where the energy is. If the user leads with a mechanic, start at Systems. If they describe a feeling, start at Target Feeling. If they mention a reference game, start at Genre & Style.
- Drill, don't move on. When the user gives a vague or high-level answer, stay on that thread. Ask the follow-up that forces them to be specific. See "Drilling into Vague Ideas" above — this is where most of the value comes from.
- Dive when something is interesting. A throwaway answer about "the world is post-apocalyptic" might open a rich thread about Narrative & World, Economy (scarcity), or Fantasy & Role. Follow it.
- Ask open-ended, resolve with options. Most questions should be free-form
dialogue — the user types their thoughts. Use
AskUserQuestiononly when you have 2-3 concrete design proposals and need the user to choose among them (e.g. picking a core loop direction, choosing between two system architectures). - Spawn
gat-designermid-interview when you need to synthesize answers into a concrete proposal (core loop draft, system list, trade-off analysis). Present what the agent returns, then ask about it. - Spawn
gat-artistmid-interview when the visual identity is too vague or conflicting. Ask for a concise art-direction proposal: references, palette, shape language, readability priorities, asset groups, and production limits. Present the proposal, then ask the user what to keep or change. - Loop back naturally. If a later answer contradicts an earlier assumption, point it out and resolve the tension. Don't pretend consistency exists when it doesn't.
- Know when to stop. The interview has covered enough when:
- The core loop is clear and the user can describe it in their own words
- The system list is named with rough dependencies
- The global visual direction has references, palette or mood, readability priorities, and asset group strategy
- Scope boundaries are set
- The user starts repeating themselves rather than adding new information
Phase 3: Write or Summarize
If Mode is design
Before writing, summarize what's been decided across gameplay, systems, scope, and visual direction. Ask:
"Ready to write the design docs?" Options:
Yes, write them/Let me keep discussing
If yes, read templates:
../../assets/templates/design/game-overview.md../../assets/templates/design/systems-index.md../../assets/templates/design/global-art.md
Step 1 — Spawn gat-designer to write both foundational design files in one pass:
gat/overview/game.mdgat/overview/systems-index.md
Pass all interview answers, the confirmed system list with dependencies, and the game overview and systems index templates.
Instruct the designer to populate the Key Design Decisions section in game.md:
record each foundational choice as a short paragraph, and add a Why: note when
the rationale or rejected alternatives need to be stated — drawing from the interview's
tension-resolution moments.
Step 2 — Spawn gat-artist to write the global art direction:
gat/overview/art-direction.md
Pass all interview answers, gat/overview/game.md, gat/overview/systems-index.md, the confirmed visual
identity decisions, and the global-art template. The artist should establish the
shared visual identity, references, palette, readability rules, asset groups,
production standards, and constraints for later system art docs.
If Mode is discuss
Summarize what was decided and what remains open. No files written.
Suggest:
$gat-brainstorm(withoutdiscuss) to turn this discussion into foundational design docs and global art direction$gat-storyif the game needs narrative (after foundational design docs exist)$gat-milestoneto plan milestone slices (after foundational design docs, and narrative if needed, exist)
Phase 4: Hand Off
Summarize what was created or discussed.
If design docs were written, suggest next steps:
$gat-storyif the game needs story, lore, worldbuilding, characters, quests, dialogue, or authored narrative content$gat-milestoneto break the game into milestone slices (milestone planning runs BEFORE per-system design; do NOT run$gat-designdirectly from here)
