Tv App Development

Dispatch Workflow

1. Classify the user request by primary outcome.
2. Select one primary sub-skill and only required secondary sub-skills.
3. Execute in dependency order using the sequencing rules below.
4. Deliver implementation plus validation artifacts.
5. If behavior crosses screens or states, add certification-aware regression checks.

Sequencing Rules

• Build in this order when creating new features:

1. ten-foot-information-architecture
2. spatial-navigation-engineering
3. input-abstraction-layer
4. Domain feature skill (tv-search-ux, auth-session-resilience, or personalization-upsell-decisioning)
5. leanback-data-layer-robustness
6. tv-performance-profiling
7. certification-aware-qa

• For bugfixes, start with the failing behavior's owning skill, then run:

1. tv-performance-profiling if latency/jank is involved
2. certification-aware-qa for regression containment

Trigger Map

• Focus loss, trapped navigation, wrong directional moves: Use spatial-navigation-engineering
• Keycode mismatches, repeated inputs, inconsistent Back/Menu events: Use input-abstraction-layer
• Overcrowded layouts, distant readability issues, too many clicks: Use ten-foot-information-architecture
• TV keyboard, IME, remote typing latency, suggestions/voice hooks: Use tv-search-ux
• QR/device code login, token expiry loops, idle timeout recovery: Use auth-session-resilience
• Remote-friendly upsell prompts, dismissal/revisit flows, funnel events: Use personalization-upsell-decisioning
• Stalling rails, weak loading states, retry/backoff, focus-safe errors: Use leanback-data-layer-robustness
• Slow startup, rail scroll jank, decode pressure, memory churn: Use tv-performance-profiling
• Submission readiness, cross-platform edge-case coverage: Use certification-aware-qa

Non-Negotiable Product Rules

• Keep D-pad navigation deterministic per screen state.
• Enforce predictable Back behavior for modals, routes, and app exit.
• Preserve or safely recover focus after async updates and resume.
• Ensure every critical user flow is fully operable without pointer input.

Output Contract

For multi-skill tasks, return:

1. Active skills header line.
2. Scope and selected sub-skills.
3. Ordered implementation plan.
4. Concrete code and config changes.
5. Validation evidence (tests, traces, or matrix rows).
6. Residual risks and next checks.

References

• Use references/dispatch-matrix.md for request-to-skill mapping and default bundles.