feat(dashboards): DashboardApp factory + all 8 dashboard migrations

.claude-plugin/marketplace.json

@@ -1,7 +1,7 @@
 {
     "$schema": "https://anthropic.com/claude-code/marketplace.schema.json",
     "name": "genesis-tools",
-    "version": "1.0.31",
+    "version": "1.0.32",
     "description": "Plugins for GenesisTools CLI development and management",
     "owner": {
         "name": "genesiscz",
@@ -11,7 +11,7 @@
         {
             "name": "genesis-tools",
             "description": "Skills and utilities for working with GenesisTools CLI toolkit. Provides guidance for discovering, executing, and troubleshooting genesis tools with integrated workflow support.",
-            "version": "1.0.31",
+            "version": "1.0.32",
             "author": {
                 "name": "GenesisTools"
             },

README.md

@@ -66,7 +66,7 @@ Commands are invoked manually with `/gt:<name>`:
 | [`gt:automate`](#gtautomate) | Build or run multi-step `tools` CLI automation presets |
 | [`gt:timelog`](#gttimelog) | Sync Timely → Azure DevOps timelogs and fill Clarity PPM timesheets |
 | [`gt:claude-history`](#gtclaude-history-command) | Search conversation history by keywords, files, commits, or time range |
-| [`gt:question`](#gtquestion) | Answer-only mode — research and explain without modifying code |
+| [`gt:question`](#gtquestion) | Answer a question and preserve the Q→A to the local question store for later review |
 
 ### Skills
 
@@ -83,7 +83,6 @@ Skills activate automatically when you mention relevant topics in conversation:
 | [`gt:typescript-error-fixer`](#gttypescript-error-fixer) | "fix type errors", "eliminate any types" | Systematic 4-phase TS error fixing with zero `any` tolerance |
 | [`gt:git-rebaser`](#gtgit-rebaser) | "rebase branches", "cascade rebase", "update child branches" | Guided rebase cascade for branch hierarchies with `--onto` |
 | [`gt:living-docs`](#gtliving-docs) | "bootstrap docs", "validate documentation", "audit docs" | Self-maintaining documentation system with context rules |
-| [`gt:writing-plans`](#gtwriting-plans) | "write a plan", "create implementation plan" | Write bite-sized TDD implementation plans with exact file paths |
 | [`debugging-master`](#debugging-master) | "debug runtime", "why is this slow", "add logging" | Hypothesis-driven runtime debugging with instrumentation toolkit |
 
 ---
@@ -152,14 +151,14 @@ Supports keyword search, file-based search, commit-based search, tool filtering,
 
 #### `gt:question`
 
-Answer-only mode — Claude researches your question (reads files, searches code, browses the web) but **never modifies any files**.
+Answer a question, then preserve the Q→A to the local question store so it can be reviewed later (`tools question log` / `tools question tail`). Also fires automatically right after you answer an important question interjected mid-session.
 
 ```bash
 /gt:question "how does the caching layer work?"
 /gt:question "what's the difference between timelog add and prepare-import?"
 ```
 
-Includes a refinement loop: after each answer, asks if you want it simpler, longer, shorter, or if it's good.
+Captured via the `question_answer` MCP tool, falling back to the `tools question record` CLI.
 
 ---
 
@@ -350,21 +349,6 @@ Philosophy: docs are a search index, not a textbook. Only document what can't be
 
 </details>
 
-#### `gt:writing-plans`
-
-Write comprehensive TDD implementation plans with bite-sized tasks, exact file paths, and complete code.
-
-<details>
-<summary><b>What you get</b></summary>
-
-- Each step is one action (2-5 minutes): write failing test → verify fail → implement → verify pass → commit
-- Exact file paths, exact commands with expected output
-- Execution handoff: subagent-driven (this session) or parallel session
-
-Plans are saved to `.claude/plans/YYYY-MM-DD-<feature-name>.md`.
-
-</details>
-
 #### `debugging-master`
 
 Hypothesis-driven runtime debugging with a structured instrumentation toolkit. Instrument code → reproduce → analyze real data.

package.json

@@ -23,6 +23,7 @@
     "devDependencies": {
         "@anthropic-ai/claude-code": "^2.1.45",
         "@biomejs/biome": "^2.3.12",
+        "@playwright/test": "1.57.0",
         "@radix-ui/react-dialog": "^1.1.15",
         "@radix-ui/react-scroll-area": "^1.2.10",
         "@radix-ui/react-slot": "^1.2.4",
@@ -58,6 +59,7 @@
         "react-devtools-core": "^7.0.0",
         "tailwind-merge": "^3.5.0",
         "ts-morph": "^28.0.0",
+        "tw-animate-css": "^1.3.6",
         "vite-tsconfig-paths": "^6.1.1",
         "web-vitals": "^5.2.0"
     },

plugins/genesis-tools/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
     "name": "genesis-tools",
-    "version": "1.0.31",
+    "version": "1.0.32",
     "description": "Skills and utilities for working with GenesisTools CLI toolkit. Provides guidance for discovering, executing, and troubleshooting genesis tools with integrated workflow support.",
     "author": {
         "name": "GenesisTools"

plugins/genesis-tools/skills/question/SKILL.md

@@ -1,43 +1,34 @@
 ---
 name: question
-description: Answer-only mode — research and explain without modifying code. Use when the user wants an explanation/rationale ("why did you choose X", "how does Y work", "what are the tradeoffs"), not an implementation. Invocable as /question.
+description: Answer a question and preserve the Q→A for later review. Use when the user runs `/question <x>` (answer it, then save), or right after you give a substantive answer worth keeping to a question interjected mid-session ("why X over Y", "how does Y work", decisions, rationale, tradeoffs). Not for routine instructions you execute or quick status checks/acknowledgements ("ok", "thanks", "pushed yet?").
 argument-hint: "<your question>"
-allowed-tools:
-    - Read
-    - Glob
-    - Grep
-    - LSP
-    - Task
-    - AskUserQuestion
-    - WebFetch
-    - WebSearch
-    - ToolSearch
-    - mcp__genesis-tools__question_answer
-    - Bash(read-only commands only — no file modifications)
 ---
 
-# Question — Answer-Only Mode
+# Question — answer & preserve
 
-The user asked: **$ARGUMENTS**
+This fires two ways. Branch on whether `$ARGUMENTS` is present:
 
-You are in **answer-only mode**. Your ONLY job is to answer the question above. You MUST NOT modify/write/create files, propose code changes, say "I'll fix this", use Write/Edit/NotebookEdit, or create plans/TODOs.
+**A. Direct invocation — `$ARGUMENTS` is set.** The user asked: **$ARGUMENTS**
+Research if needed (read/search code, look things up), then give a clear, focused, self-contained answer — rationale, links, refs as warranted. Then preserve it (see **Log**).
 
-You MAY: read & search code (Read/Glob/Grep/LSP), Task(Explore) for research, web search, read-only Bash, and the `question_answer` MCP tool (see "Logging" — its persistence runs in the MCP server process, not you editing files; this does not violate answer-only mode).
+**B. Auto-trigger — `$ARGUMENTS` is empty.** You just answered a substantive question/directive/nudge the user interjected while you were doing other work. Capture that Q→A so they can review it later without scrolling back, then return to your prior work.
 
-## Workflow
+## Log — the whole point
 
-1. **Research if needed** — read files, search code, look things up.
-2. **Give a clear, focused answer** — match depth to complexity; complete and self-contained (rationale, links, refs, code snippets as warranted). This is NOT a lossy summary.
-3. **Refinement loop** — ask via AskUserQuestion:
+Call the `question_answer` MCP tool ONCE with:
 
-   - "Good answer, thanks" → settle (go to step 4).
-   - "Explain like I'm a junior" / "Longer" / "Shorter" → rewrite per selection, ask again.
+- `question`: the user's question, verbatim or lightly cleaned.
+- `answer`: your **complete** answer in markdown (the real answer — rationale, links, refs — not a lossy one-liner).
+- `tag`: `question` (asking why/what/how), `directive` (told you to do/decide something), or `action` (a "did you / should you" nudge you acted on — include the result, e.g. "pushed @ abc1234, CI green").
+- `refs`: optional commits/files/URLs you referenced.
+- `agentLabel`: if you are a subagent, your role/task in 2–4 words.
 
-4. **MANDATORY closing step — log the final answer.** Unless the user passed `--no-log` in `$ARGUMENTS`, after the answer settles call the `question_answer` MCP tool ONCE with: `question` = the user's question (lightly cleaned), `answer` = your **final refined** answer in full markdown, `tag` = `question` (or `directive` if it was an instruction-shaped ask), `refs` = any commits/files/URLs cited. Then tell the user one line: "Logged ✓ (`<id>`)" or, if the tool returns a sink error, relay its `remedy` verbatim. If the `question_answer` tool is unavailable, fall back to `Bash: tools question record --q "…" --a-file <tmp> --tag question`. Do not skip this step; it is the whole point of the mode.
+If the `question_answer` MCP tool is unavailable, fall back to the CLI:
 
-## Answer Style
+```bash
+tools question record --q "<question>" --a-file <tmp-file-with-answer> --tag <tag>
+```
 
-Default mid-length technical; junior = analogies/no jargon; longer = edge cases/examples; shorter = bullets only; combined applies both.
-
-<!-- Codex parity (verified 2026-05-18): this same skill is invokable as /question in Codex (user-confirmed Codex supports same-name skill invocation); the question_answer MCP server is registered for Codex via `tools claude mcp install --agent codex` (mcp-manager CodexProvider → ~/.codex/config.toml [mcp_servers.genesis-tools]). No separate ~/.codex/prompts/question.md fallback needed. -->
+Then tell the user one line — "Logged ✓ (`<id>`)" — or, if the tool returns a sink error, relay its `remedy` verbatim.
 
+**Skip logging** if: the interjection was pure acknowledgement ("ok", "thanks", "continue"); there was no substantive answer to capture; or the user passed `--no-log` in `$ARGUMENTS`.