# Game Direction

> Use when game work starts without a stable direction: a creator asking "what game should I make", turning taste or references into a concept and creative brief, adopting an existing direction or pitch, judging "is this idea worth building" or whether to continue or pivot or kill an existing project, reviewing whether direction artifacts are stable enough for production, or auditing an ambiguous game repo, prototype, or notes. NOT for milestone planning or implementation (use game-production) and NOT for reviewing builds or readiness claims (use game-review).

- Skill: `gigio1023/game-direction` (Agent Skill, multi-file: 4 files)
- Install (CLI): `npx skillmds@latest add gigio1023/game-direction`
- Raw SKILL.md: https://api.skillmd.com/api/skills/gigio1023/game-direction/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: gigio1023 (https://skillmd.com/u/gigio1023)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/gigio1023/game-direction

---


# Game Direction

Turn creator intent into a direction the creator recognizes as theirs: a short
creative brief plus one named first playable build. Direction becomes stable
through play, not through longer prose, so every route here ends with a
playable build target, never with documentation alone.

## Route First

Classify the entry before doing direction work:

| Mode | When | First move |
| --- | --- | --- |
| `direction-carry-in` | A thesis, pitch, mandate, prototype, or public promise already exists | Preserve it. Audit what is still unproven. Do not rewrite direction because an artifact is incomplete. |
| `taste-to-game` | The creator has taste, references, desired feelings, or hard noes, but no stable direction | Elicit taste, then build a small concept slate. Read `references/taste-to-game.md`. |
| `brownfield-unknown` | A repo, notes, or assets exist but authority and direction are unclear | Audit before writing anything new: separate playable truth from claims in notes, record who owns direction, then reroute to one of the modes above. |

Two deeper tools:

- When the creator needs a real conversation — philosophy, player role,
  whole-game frame, first playable build — run the interview in
  `references/creator-interview.md`.
- When judging whether a thesis, pillars, player-experience target, or concept
  slate is solid, use the checklists in `references/direction-review.md`.

## Authority

The human owns taste, concept selection, Direction Lock (the explicit decision
that fixes direction), and public promises. You may ask, propose, compare,
recommend rejection, draft candidate briefs, and plan the first playable build.
When the next step needs one of the human-owned calls, stop and return
`USER_DECISION_NEEDED` with the exact decision required — do not choose for
them, and do not present your own inference as their taste.

An audit or review request authorizes inspection and findings, not rewriting
project files or locking direction. A request to create or update direction
artifacts authorizes reversible local drafts; keep them candidate material
until the creator accepts them. Publishing a promise or making a destructive
change still requires explicit approval.

## Concept Status

Track where the project is with one status:
`UNROUTED`, `TASTE_DISCOVERY`, `CONCEPT_SLATE`, `CREATIVE_BRIEF`,
`FIRST_BUILD_PLANNED`, `DIRECTION_LOCKED`, or `PARKED`. A concept slate or
candidate brief is not active scope until the creator selects it; parked and
rejected concepts keep their reasons. Always report one exact value from this
set; prose such as "unchanged" is not a status. Naming a build for an accepted
carry-in can produce `FIRST_BUILD_PLANNED`, but never implies Direction Lock.

## Output

End a routing or discovery pass with:

```text
Entry mode:
Direction source:
Authority owner:
Concept status:
Creative brief or candidate brief:
First playable build:
Safe assumptions made:
Human decision needed:
Next step:
```

Preserve every field; write `none` when one does not apply. Tag load-bearing
facts by source: user fact, repo fact, safe assumption, or blocker. Do not close
direction work while a blocker remains.

## Gotchas

- Do not ask a novice creator for genre, core loop, engine, or monetization
  before taste has been elicited; ask for remembered moments and desired
  feelings first.
- Do not generate idea catalogs. Three to five comparable, rejectable concepts
  beat twenty.
- Do not smooth away contradictions in the creator's taste too early; surface
  one contradiction at a time.
- Do not turn discovery into documentation work: if the session produces prose
  but no named playable build, the route is not done.
- If the first playable build cannot be named in one sentence (one player
  action, one visible consequence), the concept is not ready for Direction
  Lock.