Game Software Requirements Specification
Use When
- A game needs a controlled SRS across client, server, content and operations.
Do Not Use When
- The concept is unbounded; use the game product brief first.
- Do not prescribe engine or service APIs without a pinned decision.
Required Inputs
| Artefact |
Source or provider |
Required? |
Behaviour when missing |
| Approved brief, journeys, target matrix, rules and risks |
Product, design, engineering and production |
Yes |
Stop affected requirements and issue a gap record. |
| Online, economy/ads, learning, wellbeing, data, safety, culture/rights, visual and platform obligations |
Specialist owners |
Conditional |
Exclude or mark the feature blocked. |
Workflow
- Establish boundaries across client, authoritative services, platform services, content/build pipeline and operations.
- Assign
GREQ-* and specify action, precondition, authoritative transition, feedback, failure and oracle.
- Specify saves, time/randomness, input, content, gameplay accessibility, localisation, age treatment, learning evidence limits, wellbeing/stopping, and fact/fiction/consultation/permission controls.
- For online play specify authority, sessions, replication intent, reconnect, version skew and abuse cases.
- Add conditioned device performance, memory, loading, thermal, network and storage measures.
- Specify telemetry purpose, economy integrity, privacy by SDK, build provenance, rollback and ethical commercial behaviour. If ads are in scope, require clear identification, a predictable genuine break after the requested action, neutral decline/close, cap/pace, no-fill/offline/error and kill switch; prohibit interruption of play, learning, narrative, saving, recovery, stopping and exit.
- Specify visual-direction and UI acceptance as observable game outcomes and evidence while deferring craft to the five game skills in
design-system-skills.
- Stop on an ownerless transition, unresolved child/cultural/right boundary or unverifiable threshold; recover with a decision request.
Outputs
| Artefact |
Consumer |
Acceptance condition |
| Game SRS and register |
Architecture, implementation, QA and operations |
Every behaviour has owner, boundary, condition, oracle and trace. |
Evidence Produced
| Evidence |
Reviewer |
Acceptance condition |
| Requirement-source-oracle trace |
Requirements reviewer |
No requirement rests only on assumption or editor behaviour. |
Capability and permission boundaries
Read and search are allowed. Editing baselines requires authority; paid services, policy or certification require approval.
Degraded mode
Return the narrowest qualified subsystem, mark unavailable evidence not assessed, and block dependent claims.
Decision Rules
| Choice or condition |
Action |
Failure or risk avoided |
| Mutable field has no authority |
Block and assign ownership. |
Desynchronisation or exploit. |
| NFR lacks conditions |
Replace or mark provisional. |
Decorative requirements. |
| Trust/money boundary exists |
Add validation, idempotency and recovery. |
Fraud or lost state. |
Quality Standards
- Use stimulus, condition, response, measure, tolerance and verification method.
Anti-Patterns
- “Fun” as requirement. Fix: define research evidence.
- “Runs smoothly.” Fix: name hardware, build and metric.
- Client-authoritative economy. Fix: trusted validation.
- Save migration omitted. Fix: add corruption and interruption cases.
- Menu-only accessibility. Fix: cover gameplay verbs.
- Retention or ad revenue as a solitary acceptance measure. Fix: pair with agency, trust, wellbeing, privacy, fairness and stopping evidence.
References