docs(kano): Kano-model prioritization of the open backlog + survey + label scheme

[Lykhoyda]

What

Adds docs/kano-model/ — a deep-research deliverable applying the Kano model to prioritize rn-dev-agent's 16 open issues, plus reusable tooling.

File	Contents
2026-06-04-kano-model-research-report.md	Cited report: origins (Kano 1984), the 5 categories, the functional/dysfunctional evaluation table, SI/DI coefficients, the no-survey proxy adaptation (arXiv 2303.03798, BERT @0.928), critiques (Google Research; N≥200 reliability), and the Kubernetes-derived OSS triage model.
kano-backlog-categorization.md	All 16 issues categorized + Kano-gated priority waves, factoring in that #202/#186/#201 largely shipped (#218/#188).
kano-survey-template.md	Functional/dysfunctional questionnaire tailored to the plugin's features.
kano-label-scheme-and-triage.md	kano:* + effort:* labels, ready gh commands, monthly re-triage cadence.

Applied

The kano:* labels from the scheme have been created and applied to all 16 open issues (Must-be ×6, Performance ×7, Attractive ×2, Indifferent ×1).

Priority order (Kano-gated):

• Wave 1 — Must-be, open: bug: CDP connection wedges ("Already connecting") + misleading "Metro not found" when Metro is up with 0 Hermes targets #208 → bug: CDP MCP fails with -32000 when an orphaned bridge from a prior session holds the single-instance lock #182 → bug: device_* fails "rn-fast-runner not started" while cdp_status reports connected — no device-session visibility #210 → bug: native text-input unreliable (maestro inputText char-drop + device_fill fallthrough) — make cdp_interact typeText the default for text #191
• Wave 2 — Performance, cheap: bug: cdp_network_log returns duplicate entries per request (CDP Network domain + injected hook both capture) #214 · bug: cdp_mmkv delete fails on Nitro MMKV v3 (API is remove, not delete) — blocks first-class storage/auth reset #209 · bug: /observe device section — screenshot not refreshing and route out of sync #206 · enhancement: maestro_run — structured step results, partial progress on timeout, iOS clearState support #211
• Wave 3 — Attractive: enhancement: route-triggered capture for transient screens + auto-recover after many reloads (NativeWind) #212 · Strategic: CLI surface for L3 actions — bin/rn-action list/run for non-LLM consumers #108
• Wave 4 — defer/decompose: CLAUDE-MD-TEMPLATE: native-log-first error-recovery row for JS launch assertions (+ clearState/observe/known-good-replay gotchas) #199 · Session feedback (IX-2997, ~6h): 5 wins + 5 friction items #173
• Verify & close (shipped): architecture: 3-layer device control (introspection/interaction/flow) + Session Arbiter to end multi-runner foreground contention #202 · bug: maestro_run can't pass --app-file, so clearState:true flows on iOS force a raw-CLI escape #201 · iOS verification UX: stale sessions, agent-device/rn-fast-runner conflict, destructive clearState, and recovery loops #194 · enhancement: rn-dev-agent + maestro-mcp interop — driver conflict, connection persistence, runFlow allowlist, flow-drift #186

Provenance caveat

Seeded by the deep-research workflow (20 sources, 88 claims). Its verify stage hit a StructuredOutput tooling failure — every verifier abstained (0-0), so the harness's refute-by-default killed all claims (false negative). The cited facts are therefore manually vetted (canonical Kano), not machine-verified, and the report says so up front. The inferred categories are hypotheses, not measurements (no survey run); the included survey template + cadence are the path to correct them with real signal.

🤖 Generated with Claude Code