jdi-cli 0.1.0 → 0.1.1

package/core/agents/jdi-bootstrap.md

@@ -110,20 +110,59 @@ Pass `adopted=$ADOPTED` and `boundary_commit=$BOUNDARY` to the architect in Step
 
 ### Step 2.7: Multi-stack? (multi-specialist support)
 
-AskUserQuestion:
+**MANDATORY step. Never skip — even if PROJECT.md only mentions 1 language, ASK the user.**
+
+#### Step 2.7a: Auto-detect fullstack from PROJECT.md
+
+Before asking, parse `.jdi/PROJECT.md` Stack/Frameworks/Vision sections. Detect dual-stack patterns:
+
+**Backend keywords (case-insensitive):**
+`C#|.NET|dotnet|ASP\.NET|Java|Spring|Kotlin|Go|Rust|Python|Django|Flask|FastAPI|Node|Express|NestJS|Ruby|Rails|PHP|Laravel|Elixir|Phoenix`
+
+**Frontend keywords (case-insensitive):**
+`React|Vue|Svelte|Angular|Next\.?js|Nuxt|Remix|SvelteKit|Astro|Solid|Qwik|Preact|Blazor`
+
+**Mobile keywords:**
+`iOS|Swift|Android|Kotlin Mobile|React Native|Flutter|Dart|Xamarin|MAUI`
+
+**Infra keywords:**
+`Terraform|Pulumi|CloudFormation|Kubernetes|Helm|Ansible`
+
+If ≥2 categories match (e.g. backend + frontend), set `SUGGESTED_COUNT=2` and `SUGGESTED_PAIRS` accordingly.
+
+Examples of detection result from `Stack: "C# 10 + React 19"`:
+- Match: `C#` (backend) + `React` (frontend) → SUGGESTED_COUNT=2
+- Suggested pairs:
+  - `{stack_label: "Backend C#", file_glob: "**/*.{cs,csproj,sln}"}`
+  - `{stack_label: "Frontend React", file_glob: "**/*.{ts,tsx,jsx,css,scss}"}`
+
+#### Step 2.7b: AskUserQuestion (always run)
+
+If `SUGGESTED_COUNT >= 2`, the FIRST option (default-selected by AskUserQuestion) MUST be the suggested multi-stack option. Format:
+
+> "Detected fullstack project: **{detected_categories}** (e.g. {match_keywords}).
+>  Stack count?"
+>
+> Options (when SUGGESTED_COUNT=2):
+> - [Multi (2 pairs — backend + frontend) **(Recommended)**]
+> - [Single (1 specialist pair)]
+> - [Multi (3 pairs)]
+> - [Multi (custom count)]
+
+If `SUGGESTED_COUNT=1` (single language detected, e.g. just Python or just Go):
 
 > "Project stack count?
 >  - **Single-stack:** 1 doer + 1 reviewer (90% of projects)
->  - **Multi-stack:** multiple doer/reviewer pairs, each owning a file glob (fullstack: backend + frontend, mobile: iOS + Android, etc.)"
+>  - **Multi-stack:** multiple pairs (fullstack, mobile iOS+Android, infra+app, etc.)"
 >
 > Options:
-> - [Single (1 specialist pair)]
-> - [Multi (2 pairs — e.g. backend + frontend)]
-> - [Multi (3 pairs — e.g. backend + frontend + infra)]
+> - [Single (1 specialist pair) **(Recommended)**]
+> - [Multi (2 pairs)]
+> - [Multi (3 pairs)]
 > - [Multi (custom count)]
 
 If single: `SPECIALIST_COUNT=1`. Standard flow.
-If multi: `SPECIALIST_COUNT=N`. Architect loops S1-S8 N times.
+If multi: `SPECIALIST_COUNT=N`. Architect loops S1-S8 N times. Pre-fill `stack_label` + `file_glob` from `SUGGESTED_PAIRS` when available (user can edit each).
 
 For multi-stack, ask glob+label per specialist BEFORE architect S1:
 
@@ -226,7 +265,7 @@ Bootstrap ok. Next: /jdi-discuss 1
 - Never create specialist without PROJECT.md present
 - Never skip architect — bootstrap is wrapper, not generator
 - Never commit if architect returned cancelled/failed
-- 1 doer + 1 reviewer per project (default). Multi-stack = future feature
+- Single-stack (1 doer + 1 reviewer) is the default. Multi-stack (N pairs with file-glob routing) is opt-in via S2.7 — ALWAYS execute S2.7 (do not skip the question)
 </rules>
 
 <fallbacks>

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "jdi-cli",
-  "version": "0.1.0",
+  "version": "0.1.1",
   "description": "JDI (Just Do It) — lean workflow toolkit for Claude Code, GitHub Copilot, Antigravity, and OpenCode. 10 commands (7 loop + ralph + adopt + meta), 6 core agents + N per-project specialists with file-glob routing. Optional Playwright MCP + Caveman plugin install.",
   "bin": {
     "jdi": "bin/jdi.js"

package/runtimes/antigravity/skills/clean-code/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: clean-code
 description: Clean Code. Human-readable code, optimized for reading (read 10x more than written). Names reveal intent, small functions, no redundant comments, explicit error handling, no magic numbers. Applies in any language.
 triggers:

package/runtimes/antigravity/skills/dry/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: dry
 description: DRY (Don't Repeat Yourself). 1 source of truth per piece of knowledge. Detects real duplication (same decision in 2+ places) and separates from apparent duplication (same code, different reasons). Applies in any language.
 triggers:

package/runtimes/antigravity/skills/frontend-rules/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: frontend-rules
 description: Universal UI/UX and accessibility rules for any web interface. Framework-agnostic - works for React, Vue, Svelte, Solid, Angular, Blazor, Razor, Twig, Jinja, ERB, Blade, and any template engine. Based on WCAG 2.2 AA, Nielsen heuristics, Material/Apple HIG.
 triggers:

package/runtimes/antigravity/skills/frontend-validator/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: frontend-validator
 description: Validates live UI via Playwright + axe-core. Detects Playwright; installs if missing (with user consent). Spawns dev server, navigates critical routes on mobile+desktop, captures console errors, network failures, a11y violations, screenshots, layout shifts. Structured JSON output for the reviewer to parse.
 triggers:

package/runtimes/antigravity/skills/jdi-adopter/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: jdi-adopter
 description: Adopt mode for brownfield projects. Scans existing repo (manifests, layout, git, docs), infers stack/code-design, confirms with user, generates PROJECT.md + ROADMAP.md with adopted=true flag. Replaces /jdi-new for projects with code already written.
 triggers:

package/runtimes/antigravity/skills/jdi-architect/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: jdi-architect
 description: Creates new JDI agents and skills. Create mode = generic agent/skill in core. Specialist mode = per-project doer/reviewer in .jdi/agents/.
 triggers:

package/runtimes/antigravity/skills/jdi-asker/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: jdi-asker
 description: Adaptive question loop to capture locked decisions before the plan. Writes CONTEXT.md.
 triggers:

package/runtimes/antigravity/skills/jdi-planner/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: jdi-planner
 description: Generates PLAN.md for the phase. Reads CONTEXT.md (from asker) + PROJECT.md, decomposes into tasks, maps files_modified, execution order. No fluff.
 triggers:

package/runtimes/antigravity/skills/jdi-researcher/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: jdi-researcher
 description: Upfront pre-roadmap research. Reads user idea, asks key questions, researches stack/domain, generates initial PROJECT.md + ROADMAP.md. Single agent instead of multiple parallel researchers to save tokens.
 triggers:

package/runtimes/antigravity/skills/kiss/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: kiss
 description: KISS (Keep It Simple, Stupid). The simplest solution that solves the problem wins. Complexity only justified by real measured pain. Each layer/abstraction must pay its own cost. Applies in any language.
 triggers:

package/runtimes/antigravity/skills/solid/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: solid
 description: SOLID. Robert C. Martin's 5 OO design principles - SRP, OCP, LSP, ISP, DIP. Applicable in any language with types/classes/interfaces (C#, Java, TS, Python, Go, Rust, Kotlin, Swift, etc). Direct summary + anti-patterns + detection heuristics.
 triggers:

package/runtimes/antigravity/skills/yagni/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: yagni
 description: YAGNI (You Aren't Gonna Need It). Build only what the current requirement asks for. Generalize after the 3rd real case, never before. Code not written is code with no bug, no maintenance cost, no pending test. Applies in any language.
 triggers:

package/runtimes/claude/agents/jdi-adopter.md

@@ -1,4 +1,4 @@
----
+---
 name: jdi-adopter
 description: Adopt mode for brownfield projects. Scans existing repo (manifests, layout, git, docs), infers stack/code-design, confirms with user, generates PROJECT.md + ROADMAP.md with adopted=true flag. Replaces /jdi-new for projects with code already written.
 model: opus

package/runtimes/claude/agents/jdi-architect.md

@@ -1,4 +1,4 @@
----
+---
 name: jdi-architect
 description: Creates new JDI agents and skills. Create mode = generic agent/skill in core. Specialist mode = per-project doer/reviewer in .jdi/agents/.
 model: opus

package/runtimes/claude/agents/jdi-asker.md

@@ -1,4 +1,4 @@
----
+---
 name: jdi-asker
 description: Adaptive question loop to capture locked decisions before the plan. Writes CONTEXT.md.
 model: sonnet

package/runtimes/claude/agents/jdi-bootstrap.md

@@ -1,4 +1,4 @@
----
+---
 name: jdi-bootstrap
 description: Fires jdi-architect in specialist mode to generate doer + reviewer per-project. Reads PROJECT.md, drives architect, validates outputs, updates routing.
 model: sonnet
@@ -76,20 +76,59 @@ Pass `adopted=$ADOPTED` and `boundary_commit=$BOUNDARY` to the architect in Step
 
 ### Step 2.7: Multi-stack? (multi-specialist support)
 
-AskUserQuestion:
+**MANDATORY step. Never skip — even if PROJECT.md only mentions 1 language, ASK the user.**
+
+#### Step 2.7a: Auto-detect fullstack from PROJECT.md
+
+Before asking, parse `.jdi/PROJECT.md` Stack/Frameworks/Vision sections. Detect dual-stack patterns:
+
+**Backend keywords (case-insensitive):**
+`C#|.NET|dotnet|ASP\.NET|Java|Spring|Kotlin|Go|Rust|Python|Django|Flask|FastAPI|Node|Express|NestJS|Ruby|Rails|PHP|Laravel|Elixir|Phoenix`
+
+**Frontend keywords (case-insensitive):**
+`React|Vue|Svelte|Angular|Next\.?js|Nuxt|Remix|SvelteKit|Astro|Solid|Qwik|Preact|Blazor`
+
+**Mobile keywords:**
+`iOS|Swift|Android|Kotlin Mobile|React Native|Flutter|Dart|Xamarin|MAUI`
+
+**Infra keywords:**
+`Terraform|Pulumi|CloudFormation|Kubernetes|Helm|Ansible`
+
+If ≥2 categories match (e.g. backend + frontend), set `SUGGESTED_COUNT=2` and `SUGGESTED_PAIRS` accordingly.
+
+Examples of detection result from `Stack: "C# 10 + React 19"`:
+- Match: `C#` (backend) + `React` (frontend) → SUGGESTED_COUNT=2
+- Suggested pairs:
+  - `{stack_label: "Backend C#", file_glob: "**/*.{cs,csproj,sln}"}`
+  - `{stack_label: "Frontend React", file_glob: "**/*.{ts,tsx,jsx,css,scss}"}`
+
+#### Step 2.7b: AskUserQuestion (always run)
+
+If `SUGGESTED_COUNT >= 2`, the FIRST option (default-selected by AskUserQuestion) MUST be the suggested multi-stack option. Format:
+
+> "Detected fullstack project: **{detected_categories}** (e.g. {match_keywords}).
+>  Stack count?"
+>
+> Options (when SUGGESTED_COUNT=2):
+> - [Multi (2 pairs — backend + frontend) **(Recommended)**]
+> - [Single (1 specialist pair)]
+> - [Multi (3 pairs)]
+> - [Multi (custom count)]
+
+If `SUGGESTED_COUNT=1` (single language detected, e.g. just Python or just Go):
 
 > "Project stack count?
 >  - **Single-stack:** 1 doer + 1 reviewer (90% of projects)
->  - **Multi-stack:** multiple doer/reviewer pairs, each owning a file glob (fullstack: backend + frontend, mobile: iOS + Android, etc.)"
+>  - **Multi-stack:** multiple pairs (fullstack, mobile iOS+Android, infra+app, etc.)"
 >
 > Options:
-> - [Single (1 specialist pair)]
-> - [Multi (2 pairs — e.g. backend + frontend)]
-> - [Multi (3 pairs — e.g. backend + frontend + infra)]
+> - [Single (1 specialist pair) **(Recommended)**]
+> - [Multi (2 pairs)]
+> - [Multi (3 pairs)]
 > - [Multi (custom count)]
 
 If single: `SPECIALIST_COUNT=1`. Standard flow.
-If multi: `SPECIALIST_COUNT=N`. Architect loops S1-S8 N times.
+If multi: `SPECIALIST_COUNT=N`. Architect loops S1-S8 N times. Pre-fill `stack_label` + `file_glob` from `SUGGESTED_PAIRS` when available (user can edit each).
 
 For multi-stack, ask glob+label per specialist BEFORE architect S1:
 
@@ -192,7 +231,7 @@ Bootstrap ok. Next: /jdi-discuss 1
 - Never create specialist without PROJECT.md present
 - Never skip architect — bootstrap is wrapper, not generator
 - Never commit if architect returned cancelled/failed
-- 1 doer + 1 reviewer per project (default). Multi-stack = future feature
+- Single-stack (1 doer + 1 reviewer) is the default. Multi-stack (N pairs with file-glob routing) is opt-in via S2.7 — ALWAYS execute S2.7 (do not skip the question)
 </rules>
 
 <fallbacks>

package/runtimes/claude/agents/jdi-planner.md

@@ -1,4 +1,4 @@
----
+---
 name: jdi-planner
 description: Generates PLAN.md for the phase. Reads CONTEXT.md (from asker) + PROJECT.md, decomposes into tasks, maps files_modified, execution order. No fluff.
 model: opus

package/runtimes/claude/agents/jdi-researcher.md

@@ -1,4 +1,4 @@
----
+---
 name: jdi-researcher
 description: Upfront pre-roadmap research. Reads user idea, asks key questions, researches stack/domain, generates initial PROJECT.md + ROADMAP.md. Single agent instead of multiple parallel researchers to save tokens.
 model: opus

package/runtimes/claude/skills/clean-code/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: clean-code
 description: Clean Code. Human-readable code, optimized for reading (read 10x more than written). Names reveal intent, small functions, no redundant comments, explicit error handling, no magic numbers. Applies in any language.
 ---

package/runtimes/claude/skills/dry/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: dry
 description: DRY (Don't Repeat Yourself). 1 source of truth per piece of knowledge. Detects real duplication (same decision in 2+ places) and separates from apparent duplication (same code, different reasons). Applies in any language.
 ---

package/runtimes/claude/skills/frontend-rules/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: frontend-rules
 description: Universal UI/UX and accessibility rules for any web interface. Framework-agnostic - works for React, Vue, Svelte, Solid, Angular, Blazor, Razor, Twig, Jinja, ERB, Blade, and any template engine. Based on WCAG 2.2 AA, Nielsen heuristics, Material/Apple HIG.
 ---

package/runtimes/claude/skills/frontend-validator/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: frontend-validator
 description: Validates live UI via Playwright + axe-core. Detects Playwright; installs if missing (with user consent). Spawns dev server, navigates critical routes on mobile+desktop, captures console errors, network failures, a11y violations, screenshots, layout shifts. Structured JSON output for the reviewer to parse.
 ---

package/runtimes/claude/skills/kiss/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: kiss
 description: KISS (Keep It Simple, Stupid). The simplest solution that solves the problem wins. Complexity only justified by real measured pain. Each layer/abstraction must pay its own cost. Applies in any language.
 ---

package/runtimes/claude/skills/solid/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: solid
 description: SOLID. Robert C. Martin's 5 OO design principles - SRP, OCP, LSP, ISP, DIP. Applicable in any language with types/classes/interfaces (C#, Java, TS, Python, Go, Rust, Kotlin, Swift, etc). Direct summary + anti-patterns + detection heuristics.
 ---

package/runtimes/claude/skills/yagni/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: yagni
 description: YAGNI (You Aren't Gonna Need It). Build only what the current requirement asks for. Generalize after the 3rd real case, never before. Code not written is code with no bug, no maintenance cost, no pending test. Applies in any language.
 ---

package/runtimes/copilot/agents/jdi-adopter.agent.md

@@ -1,4 +1,4 @@
----
+---
 name: jdi-adopter
 description: Adopt mode for brownfield projects. Scans existing repo (manifests, layout, git, docs), infers stack/code-design, confirms with user, generates PROJECT.md + ROADMAP.md with adopted=true flag. Replaces /jdi-new for projects with code already written.
 model: gpt-5

package/runtimes/copilot/agents/jdi-architect.agent.md

@@ -1,4 +1,4 @@
----
+---
 name: jdi-architect
 description: Creates new JDI agents and skills. Create mode = generic agent/skill in core. Specialist mode = per-project doer/reviewer in .jdi/agents/.
 model: gpt-5

package/runtimes/copilot/agents/jdi-asker.agent.md

@@ -1,4 +1,4 @@
----
+---
 name: jdi-asker
 description: Adaptive question loop to capture locked decisions before the plan. Writes CONTEXT.md.
 model: gpt-5

package/runtimes/copilot/agents/jdi-bootstrap.agent.md

@@ -1,4 +1,4 @@
----
+---
 name: jdi-bootstrap
 description: Fires jdi-architect in specialist mode to generate doer + reviewer per-project. Reads PROJECT.md, drives architect, validates outputs, updates routing.
 model: gpt-5
@@ -76,20 +76,59 @@ Pass `adopted=$ADOPTED` and `boundary_commit=$BOUNDARY` to the architect in Step
 
 ### Step 2.7: Multi-stack? (multi-specialist support)
 
-AskUserQuestion:
+**MANDATORY step. Never skip — even if PROJECT.md only mentions 1 language, ASK the user.**
+
+#### Step 2.7a: Auto-detect fullstack from PROJECT.md
+
+Before asking, parse `.jdi/PROJECT.md` Stack/Frameworks/Vision sections. Detect dual-stack patterns:
+
+**Backend keywords (case-insensitive):**
+`C#|.NET|dotnet|ASP\.NET|Java|Spring|Kotlin|Go|Rust|Python|Django|Flask|FastAPI|Node|Express|NestJS|Ruby|Rails|PHP|Laravel|Elixir|Phoenix`
+
+**Frontend keywords (case-insensitive):**
+`React|Vue|Svelte|Angular|Next\.?js|Nuxt|Remix|SvelteKit|Astro|Solid|Qwik|Preact|Blazor`
+
+**Mobile keywords:**
+`iOS|Swift|Android|Kotlin Mobile|React Native|Flutter|Dart|Xamarin|MAUI`
+
+**Infra keywords:**
+`Terraform|Pulumi|CloudFormation|Kubernetes|Helm|Ansible`
+
+If ≥2 categories match (e.g. backend + frontend), set `SUGGESTED_COUNT=2` and `SUGGESTED_PAIRS` accordingly.
+
+Examples of detection result from `Stack: "C# 10 + React 19"`:
+- Match: `C#` (backend) + `React` (frontend) → SUGGESTED_COUNT=2
+- Suggested pairs:
+  - `{stack_label: "Backend C#", file_glob: "**/*.{cs,csproj,sln}"}`
+  - `{stack_label: "Frontend React", file_glob: "**/*.{ts,tsx,jsx,css,scss}"}`
+
+#### Step 2.7b: AskUserQuestion (always run)
+
+If `SUGGESTED_COUNT >= 2`, the FIRST option (default-selected by AskUserQuestion) MUST be the suggested multi-stack option. Format:
+
+> "Detected fullstack project: **{detected_categories}** (e.g. {match_keywords}).
+>  Stack count?"
+>
+> Options (when SUGGESTED_COUNT=2):
+> - [Multi (2 pairs — backend + frontend) **(Recommended)**]
+> - [Single (1 specialist pair)]
+> - [Multi (3 pairs)]
+> - [Multi (custom count)]
+
+If `SUGGESTED_COUNT=1` (single language detected, e.g. just Python or just Go):
 
 > "Project stack count?
 >  - **Single-stack:** 1 doer + 1 reviewer (90% of projects)
->  - **Multi-stack:** multiple doer/reviewer pairs, each owning a file glob (fullstack: backend + frontend, mobile: iOS + Android, etc.)"
+>  - **Multi-stack:** multiple pairs (fullstack, mobile iOS+Android, infra+app, etc.)"
 >
 > Options:
-> - [Single (1 specialist pair)]
-> - [Multi (2 pairs — e.g. backend + frontend)]
-> - [Multi (3 pairs — e.g. backend + frontend + infra)]
+> - [Single (1 specialist pair) **(Recommended)**]
+> - [Multi (2 pairs)]
+> - [Multi (3 pairs)]
 > - [Multi (custom count)]
 
 If single: `SPECIALIST_COUNT=1`. Standard flow.
-If multi: `SPECIALIST_COUNT=N`. Architect loops S1-S8 N times.
+If multi: `SPECIALIST_COUNT=N`. Architect loops S1-S8 N times. Pre-fill `stack_label` + `file_glob` from `SUGGESTED_PAIRS` when available (user can edit each).
 
 For multi-stack, ask glob+label per specialist BEFORE architect S1:
 
@@ -192,7 +231,7 @@ Bootstrap ok. Next: /jdi-discuss 1
 - Never create specialist without PROJECT.md present
 - Never skip architect — bootstrap is wrapper, not generator
 - Never commit if architect returned cancelled/failed
-- 1 doer + 1 reviewer per project (default). Multi-stack = future feature
+- Single-stack (1 doer + 1 reviewer) is the default. Multi-stack (N pairs with file-glob routing) is opt-in via S2.7 — ALWAYS execute S2.7 (do not skip the question)
 </rules>
 
 <fallbacks>

package/runtimes/copilot/agents/jdi-planner.agent.md

@@ -1,4 +1,4 @@
----
+---
 name: jdi-planner
 description: Generates PLAN.md for the phase. Reads CONTEXT.md (from asker) + PROJECT.md, decomposes into tasks, maps files_modified, execution order. No fluff.
 model: gpt-5

package/runtimes/copilot/agents/jdi-researcher.agent.md

@@ -1,4 +1,4 @@
----
+---
 name: jdi-researcher
 description: Upfront pre-roadmap research. Reads user idea, asks key questions, researches stack/domain, generates initial PROJECT.md + ROADMAP.md. Single agent instead of multiple parallel researchers to save tokens.
 model: gpt-5

package/runtimes/opencode/agents/jdi-adopter.md

@@ -1,4 +1,4 @@
----
+---
 description: Adopt mode for brownfield projects. Scans existing repo (manifests, layout, git, docs), infers stack/code-design, confirms with user, generates PROJECT.md + ROADMAP.md with adopted=true flag. Replaces /jdi-new for projects with code already written.
 mode: subagent
 model: anthropic/claude-sonnet-4-20250514

package/runtimes/opencode/agents/jdi-architect.md

@@ -1,4 +1,4 @@
----
+---
 description: Creates new JDI agents and skills. Create mode = generic agent/skill in core. Specialist mode = per-project doer/reviewer in .jdi/agents/.
 ---
 

package/runtimes/opencode/agents/jdi-asker.md

@@ -1,4 +1,4 @@
----
+---
 description: Adaptive question loop to capture locked decisions before the plan. Writes CONTEXT.md.
 mode: subagent
 model: anthropic/claude-sonnet-4-20250514

package/runtimes/opencode/agents/jdi-bootstrap.md

@@ -1,4 +1,4 @@
----
+---
 description: Fires jdi-architect in specialist mode to generate doer + reviewer per-project. Reads PROJECT.md, drives architect, validates outputs, updates routing.
 mode: subagent
 model: anthropic/claude-sonnet-4-20250514
@@ -80,20 +80,59 @@ Pass `adopted=$ADOPTED` and `boundary_commit=$BOUNDARY` to the architect in Step
 
 ### Step 2.7: Multi-stack? (multi-specialist support)
 
-AskUserQuestion:
+**MANDATORY step. Never skip — even if PROJECT.md only mentions 1 language, ASK the user.**
+
+#### Step 2.7a: Auto-detect fullstack from PROJECT.md
+
+Before asking, parse `.jdi/PROJECT.md` Stack/Frameworks/Vision sections. Detect dual-stack patterns:
+
+**Backend keywords (case-insensitive):**
+`C#|.NET|dotnet|ASP\.NET|Java|Spring|Kotlin|Go|Rust|Python|Django|Flask|FastAPI|Node|Express|NestJS|Ruby|Rails|PHP|Laravel|Elixir|Phoenix`
+
+**Frontend keywords (case-insensitive):**
+`React|Vue|Svelte|Angular|Next\.?js|Nuxt|Remix|SvelteKit|Astro|Solid|Qwik|Preact|Blazor`
+
+**Mobile keywords:**
+`iOS|Swift|Android|Kotlin Mobile|React Native|Flutter|Dart|Xamarin|MAUI`
+
+**Infra keywords:**
+`Terraform|Pulumi|CloudFormation|Kubernetes|Helm|Ansible`
+
+If ≥2 categories match (e.g. backend + frontend), set `SUGGESTED_COUNT=2` and `SUGGESTED_PAIRS` accordingly.
+
+Examples of detection result from `Stack: "C# 10 + React 19"`:
+- Match: `C#` (backend) + `React` (frontend) → SUGGESTED_COUNT=2
+- Suggested pairs:
+  - `{stack_label: "Backend C#", file_glob: "**/*.{cs,csproj,sln}"}`
+  - `{stack_label: "Frontend React", file_glob: "**/*.{ts,tsx,jsx,css,scss}"}`
+
+#### Step 2.7b: AskUserQuestion (always run)
+
+If `SUGGESTED_COUNT >= 2`, the FIRST option (default-selected by AskUserQuestion) MUST be the suggested multi-stack option. Format:
+
+> "Detected fullstack project: **{detected_categories}** (e.g. {match_keywords}).
+>  Stack count?"
+>
+> Options (when SUGGESTED_COUNT=2):
+> - [Multi (2 pairs — backend + frontend) **(Recommended)**]
+> - [Single (1 specialist pair)]
+> - [Multi (3 pairs)]
+> - [Multi (custom count)]
+
+If `SUGGESTED_COUNT=1` (single language detected, e.g. just Python or just Go):
 
 > "Project stack count?
 >  - **Single-stack:** 1 doer + 1 reviewer (90% of projects)
->  - **Multi-stack:** multiple doer/reviewer pairs, each owning a file glob (fullstack: backend + frontend, mobile: iOS + Android, etc.)"
+>  - **Multi-stack:** multiple pairs (fullstack, mobile iOS+Android, infra+app, etc.)"
 >
 > Options:
-> - [Single (1 specialist pair)]
-> - [Multi (2 pairs — e.g. backend + frontend)]
-> - [Multi (3 pairs — e.g. backend + frontend + infra)]
+> - [Single (1 specialist pair) **(Recommended)**]
+> - [Multi (2 pairs)]
+> - [Multi (3 pairs)]
 > - [Multi (custom count)]
 
 If single: `SPECIALIST_COUNT=1`. Standard flow.
-If multi: `SPECIALIST_COUNT=N`. Architect loops S1-S8 N times.
+If multi: `SPECIALIST_COUNT=N`. Architect loops S1-S8 N times. Pre-fill `stack_label` + `file_glob` from `SUGGESTED_PAIRS` when available (user can edit each).
 
 For multi-stack, ask glob+label per specialist BEFORE architect S1:
 
@@ -196,7 +235,7 @@ Bootstrap ok. Next: /jdi-discuss 1
 - Never create specialist without PROJECT.md present
 - Never skip architect — bootstrap is wrapper, not generator
 - Never commit if architect returned cancelled/failed
-- 1 doer + 1 reviewer per project (default). Multi-stack = future feature
+- Single-stack (1 doer + 1 reviewer) is the default. Multi-stack (N pairs with file-glob routing) is opt-in via S2.7 — ALWAYS execute S2.7 (do not skip the question)
 </rules>
 
 <fallbacks>

package/runtimes/opencode/agents/jdi-planner.md

@@ -1,4 +1,4 @@
----
+---
 description: Generates PLAN.md for the phase. Reads CONTEXT.md (from asker) + PROJECT.md, decomposes into tasks, maps files_modified, execution order. No fluff.
 mode: subagent
 model: anthropic/claude-sonnet-4-20250514

package/runtimes/opencode/agents/jdi-researcher.md

@@ -1,4 +1,4 @@
----
+---
 description: Upfront pre-roadmap research. Reads user idea, asks key questions, researches stack/domain, generates initial PROJECT.md + ROADMAP.md. Single agent instead of multiple parallel researchers to save tokens.
 mode: subagent
 model: anthropic/claude-sonnet-4-20250514

package/runtimes/opencode/skills/clean-code/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: clean-code
 description: Clean Code. Human-readable code, optimized for reading (read 10x more than written). Names reveal intent, small functions, no redundant comments, explicit error handling, no magic numbers. Applies in any language.
 ---

package/runtimes/opencode/skills/dry/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: dry
 description: DRY (Don't Repeat Yourself). 1 source of truth per piece of knowledge. Detects real duplication (same decision in 2+ places) and separates from apparent duplication (same code, different reasons). Applies in any language.
 ---

package/runtimes/opencode/skills/frontend-rules/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: frontend-rules
 description: Universal UI/UX and accessibility rules for any web interface. Framework-agnostic - works for React, Vue, Svelte, Solid, Angular, Blazor, Razor, Twig, Jinja, ERB, Blade, and any template engine. Based on WCAG 2.2 AA, Nielsen heuristics, Material/Apple HIG.
 ---

package/runtimes/opencode/skills/frontend-validator/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: frontend-validator
 description: Validates live UI via Playwright + axe-core. Detects Playwright; installs if missing (with user consent). Spawns dev server, navigates critical routes on mobile+desktop, captures console errors, network failures, a11y violations, screenshots, layout shifts. Structured JSON output for the reviewer to parse.
 ---

package/runtimes/opencode/skills/kiss/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: kiss
 description: KISS (Keep It Simple, Stupid). The simplest solution that solves the problem wins. Complexity only justified by real measured pain. Each layer/abstraction must pay its own cost. Applies in any language.
 ---

package/runtimes/opencode/skills/solid/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: solid
 description: SOLID. Robert C. Martin's 5 OO design principles - SRP, OCP, LSP, ISP, DIP. Applicable in any language with types/classes/interfaces (C#, Java, TS, Python, Go, Rust, Kotlin, Swift, etc). Direct summary + anti-patterns + detection heuristics.
 ---

package/runtimes/opencode/skills/yagni/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: yagni
 description: YAGNI (You Aren't Gonna Need It). Build only what the current requirement asks for. Generalize after the 3rd real case, never before. Code not written is code with no bug, no maintenance cost, no pending test. Applies in any language.
 ---