Wayfinder: next new pattern admission shortlist

[Skymly]

Destination Produce a ranked Top-3 admission shortlist plus a rejection appendix for new pattern domains to add to DesignPatterns; write the locked shortlist back into docs/ROADMAP.md F3. Map ends when the shortlist is decided and ROADMAP is updated — no pattern implementation. ## Notes - Domain: Skymly DesignPatterns (.NET primitives + Roslyn generators/analyzers); exploration over avoiding ecosystem overlap. - Skills each session should consult: /grilling, /domain-modeling, /research; always read AGENTS.md, docs/ROADMAP.md F3, and docs/design/. - Locked preferences from charting: - Deliverable: Top-3 + rejection appendix (not “winner only”, not full score dump without truncation). - Scope: open cast then converge; survey five families — (1) GoF creation/structure/behavior gaps not yet shipped as domains, (2) messaging/command routing, (3) resilience/fault tolerance, (4) concurrency/coordination, (5) functional/specification-style. - New pattern domains only — existing-pattern enhancements (e.g. Composite lazy load, Decorator compose) stay on F2+/long-term backlog and are out of this shortlist. - Rubric: ROADMAP F3 three hard gates (lightweight primitive + compile-time glue; compile-time exploration value; real scenario + Sample sketch) + rank passers by compile-time synergy depth (generator / analyzer / CodeFix / DI hooks). - Final step of the map: update docs/ROADMAP.md F3 with Top-3 and rejection appendix (Docs module); still no runtime/generator code. - Language: issues/commits English; chat with maintainer in 简体中文. - Plan, don't implement patterns inside this map. ## Decisions so far ## Not yet specified - Exact relative weights inside “compile-time synergy depth” when ranking near-ties. - Quality bar for “Sample sketch” (how concrete before a candidate may pass the third hard gate). - Whether admitting a pattern also reserves diagnostic ID ranges / Design Doc stubs at admission time, or only at implementation. - How near-duplicates vs shipped domains are treated (e.g. Observer-shaped ideas vs existing Event Aggregator) once the longlist exists. - Whether Top-3 order is also the default implementation order for later milestones. ## Out of scope - Implementing any new pattern runtime, generator, analyzer, or sample project. - Ranking or admitting enhancements to already-shipped patterns. - API freeze, stable NuGet, VSIX/CompletionProvider, packaging policy changes. ## Tickets - [x] #245 — Research: inventory shipped pattern domains and F3 pool - [x] #246 — Research: GoF gap candidates for new pattern domains - [x] #247 — Research: messaging and command-routing pattern candidates - [x] #248 — Research: resilience and fault-tolerance pattern candidates - [x] #249 — Research: concurrency and coordination pattern candidates - [x] #250 — Research: functional and specification-style pattern candidates - [x] #251 — Grilling: apply hard gates and draft rejection appendix (blocked by researches above) - [x] #252 — Research: compile-time synergy depth for hard-gate passers (blocked by hard-gate grilling) - [x] #253 — Grilling: lock Top-3 admission shortlist (blocked by synergy research) - [x] #254 — Task: write Top-3 and rejection appendix into ROADMAP F3 (blocked by Top-3 lock)

[Skymly]

Destination reached: Top-3 admission shortlist locked (#253) and written into docs/ROADMAP.md F3 (#254). Remaining open fog (Sample bar, diagnostic ID reservation at admission, Top-3 as default implementation order) can be decided at implementation time.