| name | game-narrative-and-interactive-story-design |
| description | Use when designing, reviewing, or implementing interactive game stories, narrative systems, quests, characters, dialogue, environmental storytelling, branching choices, story beats, setups/payoffs, or narrative validation; use level-world-and-content-production for level assembly and content integration. |
| metadata | {"portable":true,"compatible_with":["claude-code","codex"]} |
Game Narrative and Interactive Story Design
Convert theme, audience, player fantasy, and gameplay constraints into a causal, testable narrative system whose delivery survives player agency, interruption, and production limits.
Prerequisites
Load world-class-engineering, game-design-and-experience, level-world-and-content-production, accessibility/localisation guidance, and the project's cultural research, rights, safeguarding, and content constraints.
Use When
- Defining premise, theme, characters, wants/needs, conflict, beats, quests, dialogue, choices, consequences, environmental narrative, or story-state logic.
- Diagnosing a story that is incoherent, over-exposed, disconnected from play, culturally unsafe, or impossible to validate.
- Creating narrative requirements, state models, content schemas, review gates, localisation notes, or comprehension playtests.
Do Not Use When
- The task is level layout, encounter pacing, or content integration without a narrative-system decision; use
level-world-and-content-production.
- The task is prose-only fiction with no interactive delivery or runtime state.
- Cultural fact claims lack evidence; return to research and cultural review before canonising them.
Required Inputs
| Input | Required | Missing-input response |
|---|
| Audience, player fantasy, genre/tone, platform and session shape | yes | Mark the premise provisional and stop detailed scripting. |
| Approved cultural/source ledger and rights/sensitivity constraints | conditional | Block historical or community-specific canon. |
| Gameplay loop, player verbs, level/content topology and save model | yes | Produce only a narrative hypothesis, not runtime logic. |
| Content budget, localisation/VO plan, rating and accessibility constraints | yes | Flag cost and delivery as unvalidated. |
Workflow
- State the player-facing narrative promise, theme, audience need and non-negotiable cultural boundaries in plain language.
- Build a causal spine: setup, destabilising event, dramatic question, escalating obstacles and choices, climax, resolution, and changed state. Every beat must follow from an action or consequence.
- Define characters by observable want, deeper need, stakes, tactics, contradiction, relationships, agency and change—not biography volume.
- Map the spine to player verbs, spaces, encounters and session boundaries. Choose authored, branching, environmental, systemic/emergent or hybrid delivery deliberately.
- Model story states, prerequisites, mutually exclusive branches, convergence, failure/recovery, save/load, replay and late-join behaviour. Give each state stable identifiers.
- Specify each delivery unit with trigger, payload, player control, interruption policy, fallback, localisation fields, accessibility channel and acceptance oracle.
- Track setups, reminders and payoffs; separate essential comprehension from optional depth. Never place a critical fact in one fragile channel.
- Prototype with the cheapest faithful form, then test comprehension, motivation, pacing, choice legibility and cultural interpretation with named builds and audiences.
- Stop on broken causality, unreachable state, missing cultural approval, inaccessible critical information, uncontrolled scope, or an untestable emotional claim.
- Load
references/narrative-playtest-loop.md for the cross-discipline contract. Retain
failed-path, branch, audio-off, localisation-expansion, and accessibility evidence;
standardise only a change that improves the player hypothesis without breaking
gameplay, performance, rights, or safety constraints.
Outputs
| Artefact | Acceptance condition |
|---|
| Narrative promise, causal spine and character contracts | Wants, stakes, actions, consequences and changed states are explicit. |
| Beat/quest/state map | IDs, prerequisites, branches, convergence, failure and persistence are deterministic. |
| Narrative delivery specification | Trigger, channel, interruption, fallback, localisation and accessibility are defined. |
| Setup/payoff and evidence register | Every essential setup has a reachable payoff and test evidence. |
Evidence Produced
| Category | Artifact | Format | Example |
|---|
| Correctness | Reachability/state-transition tests and setup-payoff coverage. | Test report and coverage matrix | docs/narrative/reachability-tests.md |
| UX quality | Comprehension, motivation, pacing and choice-legibility playtest results. | Playtest report with observations and findings | docs/narrative/player-experience.md |
| UX quality | Cultural safety review | Review record: named reviewer, source/provenance link, decision and unresolved caveat | docs/narrative/cultural-safety-review.md |
| Release evidence | Build/content revision, localisation completeness and known narrative defects. | Release record with revision identity, localisation status and defect list | docs/narrative/release-evidence.md |
Capability and Degraded Mode
Editing canon, dialogue, localisation source or runtime content requires task authority; community representation and public claims require accountable review. Without a playable build, return a desk-reviewed state model and test plan with runtime comprehension, pacing and interruption marked not assessed.
Decision Rules
| Condition | Action |
|---|
| Plot beat does not alter player knowledge, goal, relationship, capability or world state | Cut, combine or make its purpose explicit. |
| Critical information has one missable delivery channel | Add an accessible, diegetically appropriate recovery path. |
| Choice has no legible difference or consequence | Reframe as flavour or add a measurable consequence; do not market false agency. |
| Branch cost exceeds budget | Share states/content intentionally and preserve consequence through variables, reactions or later payoffs. |
| Historical authenticity and fun appear to conflict | Escalate to research/cultural owners; label dramatization and never invent certainty. |
Quality Standards
- Prefer causal verbs and observable player evidence over adjectives such as “immersive” or “meaningful”.
- Interactive flow outranks screenplay shape: the player must understand what to do, why, and what changed.
- Preserve player control or specify the exact, tested reason for taking it away.
- Treat the supplied story text as a design source, not a substitute for cultural evidence or project canon.
- Complete the mechanics-to-meaning matrix for consequential mechanics; agency, feedback, state transitions, and player interpretation must be evidenced rather than assumed.
- Publish the initial product audit at
min(raw_score, 65) before any improvement claim; the chosen change targets 95/100 and has a reversible test, guardrails, and re-audit evidence.
Anti-Patterns
- Lore dump before player need. Fix: reveal information through action, consequence and optional depth.
- Branching diagram without state ownership. Fix: assign stable IDs, predicates, persistence and recovery.
- Emotional acceptance criterion (“player feels heroic”). Fix: define observable comprehension/choice/behaviour evidence.
- One “representative” reviewer. Fix: identify role, community relationship, scope and unresolved disagreement.
- Cutscene logic detached from gameplay. Fix: map every major beat to player verbs and state change.
Read Next
level-world-and-content-production for spatial and encounter integration.gameplay-systems-engineering for runtime state implementation.game-testing-polish for playtest and regression evidence.
References
