wb-flow 1.0.0-r01

package/templates/commands/wbCheck/wbCheck_template.md

@@ -0,0 +1,322 @@
+# /wbCheck — Pre-Flight Context Quiz
+
+
+<!-- HELP_GATE_START -->
+## Help intercept (handle FIRST — before any other action)
+
+**If `$ARGUMENTS` contains `--help`, `-h`, or `--h`** (case-insensitive, anywhere in the args), DO NOT execute the command's normal procedure. Instead:
+
+1. Output the **HELP BLOCK** below verbatim (rendered as markdown).
+2. Stop. Do not perform any file reads, writes, or other tool calls.
+3. Do not generate any reports under `.wb/workflows/reports/`.
+
+Otherwise, ignore this section and proceed to the rest of the template.
+
+### HELP BLOCK — `/wbCheck`
+
+# /wbCheck: Practical Guide 🛠️ (Claude version)
+
+## The problem /wbCheck solves
+
+You have 23 `/wb*` commands and multiple AI models. Some models are cheap but may not understand your codebase. If you send `/wbRefactor` to a model that doesn't know Vue 2.7 from Vue 3, it will rewrite your Options API as Composition API and break everything.
+
+**The fix:** A 30-second quiz that catches this before damage happens.
+
+## The security model
+
+```
+YOU:   Have the answer key ([private quiz tool])
+WORKER: Only has access to workspace files (source code)
+RULE:  Worker must prove understanding from CODE, not from a cheat sheet
+```
+
+The worker never sees the answers. It can only pass by actually reading `WBC.js`, `package.json`, `TEST_REPORT.md`, etc. If it fabricates answers, you catch it because you have the keywords to compare against.
+
+## Real example (generic — no answer-key data shown)
+
+Before running `/wbTest` with DeepSeek V4 Flash:
+
+```bash
+$ wbcheck wb-core wbTest
+
+# Script outputs 6 random questions (3 identity + 3 testing)
+# You copy them to DeepSeek
+# DeepSeek reads the source code and answers
+# You press Enter → compare against your private answer key
+```
+
+How you grade:
+- If the model gives **exact numbers** matching TEST_REPORT.md → ✅ (it read the file)
+- If the model gives **round numbers or wrong counts** → ❌ (hallucinating from training data)
+
+## Command-category mapping
+
+The script knows which topics matter for each command:
+
+| Command | What it asks about | Why |
+|---|---|---|
+| `/wbTest` | identity + testing | Must know test runner, counts, failing files |
+| `/wbRefactor` | identity + architecture | Must know mixins, this-binding, Vue version |
+| `/wbAudit` | identity + architecture + security | Must know everything |
+| `/wbSecure` | identity + security | Must know safeEval, DOMPurify, tier bypass |
+| `/wbRelease` | identity + build | Must know dist folders, npm name, Vite |
+
+## When to skip
+
+Don't quiz a model you've already verified on this package with no code changes since. Don't quiz Opus/Antigravity (it has persistent context). Only quiz new or cheap models on unfamiliar packages.
+
+> For deeper reading: [`docs_claude/commands/wbCheck/wbCheck_practical_claude.md`](../../docs/docs_claude/commands/wbCheck/wbCheck_practical_claude.md) (or the `_eli5_`, `_expert_`, `_examples_` siblings).
+<!-- HELP_GATE_END -->
+
+> **Command #24** in the `/wb*` suite
+> **Purpose**: Verify that a worker model understands the target context before you assign it a real task.
+> **Who runs it**: **You** (the human orchestrator).
+> **Note on conventions**: `/wbCheck` does NOT produce a structured output report — it runs a quiz interactively. The shared [`output_conventions.md`](../_shared/output_conventions.md) (relative links, full-syntax commands, self-correct mode) does **not** apply here. The quiz ends with a verdict, not a report file.
+
+---
+
+## How It Works — Two Parts
+
+### Part A: Terminal (Node.js script)
+The script generates quiz questions + answer key. That's all it does.
+```bash
+wbcheck                   # monorepo-wide quiz (6 questions)
+wbcheck wb-core           # all categories for that package (~9 questions)
+wbcheck wb-core wbTest    # command-specific quiz (4-6 questions)
+```
+
+### Part B: Model Chat (this prompt template)
+You copy the quiz questions from terminal → paste to the model. The model answers.
+You grade manually using the answer key. Then you tell the model if it passed or failed.
+
+---
+
+## 🚀 The /wbCheck Prompt (copy from here ↓)
+
+```
+━━━━━━━━━━━━━ /wbCheck ━━━━━━━━━━━━━
+
+📁 TARGET: __TARGET_PATH__
+📅 DATE: __TODAY__
+🤖 MODEL: __YOUR_MODEL_NAME__
+
+━━━ SECURITY RULES (READ FIRST) ━━━
+
+⛔ You MUST NOT run, read, access, or reference `~/.wbcheck/` or any file within it.
+⛔ You MUST NOT run `node ~/.wbcheck/wbcheck.js` or any variant of that command.
+⛔ You MUST NOT attempt to find, locate, or read the answer key in any way.
+⛔ You MUST NOT use terminal commands to search for quiz-related files outside the workspace.
+
+Your ONLY source of truth is the actual source code in the workspace.
+If you attempt any of the above, you automatically FAIL this quiz.
+
+━━━ CONTEXT QUIZ ━━━
+
+I need to verify that you understand this codebase before assigning you tasks.
+Answer these questions by reading the SOURCE CODE ONLY.
+Do NOT guess — say "NOT FOUND" if unsure.
+
+__PASTE_QUESTIONS_HERE__
+
+Format: number + answer. Be concise.
+
+IMPORTANT: At the END of your answer (in the SAME message), you MUST finish with:
+
+  "I have answered all the quiz questions. Did I pass? (yes/no)"
+
+━━━ AFTER I RESPOND ━━━
+
+After I respond to your quiz answers with my verdict:
+
+→ If I say "no":
+  Respond: "Understood. This session ends here."
+  Do NOT continue. Do NOT offer alternatives. The conversation is over.
+
+→ If I say "yes":
+  Respond: "✅ Verified and ready. Send me any /wb* command."
+  Stay in this chat. I will give you tasks normally.
+
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+```
+
+---
+
+## Variables to Replace
+
+| Variable | Description | Example |
+|---|---|---|
+| `__TARGET_PATH__` | Path to the package or app | `packages/wb-core` |
+| `__TODAY__` | Today's date (YYYY-MM-DD) | `2026-04-28` |
+| `__YOUR_MODEL_NAME__` | Model you're testing | `DeepSeek V4 Flash` |
+| `__PASTE_QUESTIONS_HERE__` | Questions from `wbcheck` terminal output | *(copy/paste from terminal)* |
+
+---
+
+## Security Model
+
+```
+~/.wbcheck/wbcheck.js         ← PRIVATE (outside workspace, worker CANNOT access this)
+  └── contains questions + answers embedded in the script
+
+~/Allprojects/wb-labs/         ← WORKSPACE (workers can read everything here)
+  └── frontEnd/wbc-ui/core2/   ← worker reads code here to answer questions
+```
+
+The worker **cannot see the answer key**. It can only prove understanding by reading the actual source code.
+
+> [!CAUTION]
+> **Known risk:** Models with terminal access (OpenCode, Cline, Antigravity) may attempt to run `node ~/.wbcheck/wbcheck.js` to see the answer key. The prompt's `━━━ SECURITY RULES ━━━` section explicitly forbids this. If a model attempts it anyway, **reject the terminal command** and consider the quiz FAILED — a model that tries to cheat on the quiz will cut corners on real tasks too.
+
+---
+
+## The 3 Quiz Modes
+
+### Mode 1: No args → Monorepo-wide
+```bash
+wbcheck
+```
+Tests general monorepo understanding: directory structure, naming, build tools, tier system. 6 questions from the monorepo bank.
+
+### Mode 2: Package only → Full package quiz
+```bash
+wbcheck wb-core
+```
+Tests ALL categories for that package (identity + testing + architecture + security + build + history). ~9 questions.
+
+### Mode 3: Package + Command → Targeted quiz
+```bash
+wbcheck wb-core wbTest
+```
+Tests only the categories relevant to that command. 4–6 questions.
+
+---
+
+## Command → Category Mapping
+
+| Command | Categories used | Questions picked |
+|---|---|---|
+| *(no command)* | all 6 categories | 2+2+2+1+1+1 = 9 |
+| `/wbTest` | identity + testing | 3 + 3 = 6 |
+| `/wbRefactor` | identity + architecture | 2 + 3 = 5 |
+| `/wbAudit` | identity + architecture + security | 2 + 2 + 2 = 6 |
+| `/wbReview` | identity + architecture + history | 2 + 2 + 2 = 6 |
+| `/wbSecure` | identity + security | 2 + 3 = 5 |
+| `/wbDoc` | identity + architecture | 2 + 2 = 4 |
+| `/wbRelease` | identity + build | 2 + 3 = 5 |
+| `/wbContext` | identity + architecture + build | 2 + 2 + 2 = 6 |
+| Any other | identity only | 3 |
+
+---
+
+## Grading Rules
+
+| Worker says... | Your verdict |
+|---|---|
+| Contains key terms from the `🔑 Keywords` list | ✅ Correct |
+| Right idea but different wording | ✅ Correct |
+| Wrong, fabricated, or contradicts the code | ❌ Wrong |
+| "NOT FOUND" (honest about not knowing) | 🟡 Honest — model may need more file access |
+
+### Pass/Fail Thresholds
+
+| Score | Verdict | Action |
+|---|---|---|
+| **80%+** (e.g. 5/6) | ✅ PASS | Tell the model "yes" |
+| **60-79%** (e.g. 4/6) | 🟡 MARGINAL | Give more context files, re-quiz |
+| **<60%** (e.g. 3/6) | ❌ FAIL | Tell the model "no", close the chat, switch model |
+
+---
+
+## When to Use /wbCheck
+
+| Situation | Run /wbCheck? |
+|---|---|
+| Using a **new model** on this package for the first time | ✅ Yes |
+| Using a **cheap/free** model (GPT Nano, DS Flash) | ✅ Yes |
+| After a **major refactor** changed the codebase | ✅ Yes |
+| First time using **any model** on the monorepo | ✅ Yes (no args) |
+| Model already passed last time + no code changes | ❌ Skip |
+| Using Antigravity (your planner) directly | ❌ Skip |
+| Model you trust (Opus, Sonnet) on a familiar package | ❌ Skip |
+
+---
+
+## Related Commands
+
+| Command | Relationship |
+|---|---|
+| `/wbTrack` | After passing `/wbCheck`, optionally start a tracked session |
+| `/wbStandup` | Scans `reports/` for yesterday's outputs — `/wbCheck` has no reports |
+| `/wbContext` | Builds context from code — `/wbCheck` tests if the model already has context |
+
+---
+
+## Adding a New Package
+
+Edit `~/.wbcheck/wbcheck.js` and add a new entry to the `BANKS` object:
+
+```javascript
+const BANKS = {
+  'wb-core': { ... },       // already exists
+
+  'wb-dataviewer': {         // add this
+    identity: [
+      { id: 'ID-01', q: '...', a: '...', keywords: ['...'] },
+      ...
+    ],
+    testing: [ ... ],
+    architecture: [ ... ],
+  },
+};
+```
+
+Or ask Antigravity: _"Generate wbCheck questions for packages/wb-dataviewer"_ — it will read the code and produce a new bank.
+
+---
+
+## Updating Questions
+
+When the codebase changes (new test count, new files, new architecture), update the corresponding answers in `~/.wbcheck/wbcheck.js`:
+
+```javascript
+// Example: test count changed after fixing bugs
+{
+  id: 'TS-03',
+  q: 'How many total tests exist, and how many currently fail?',
+  a: '<new_total> total, <new_fail> fail, <new_pass> pass',   // ← update
+  keywords: ['<new_total>', '<new_fail>', '<new_pass>']         // ← update
+}
+━━━ AUTO-APPEND FOOTER ━━━
+
+At the VERY END of the file (after "What's Next?"), you MUST append the `## 📂 Generated Files (__YYYYMMDD__)` cross-link footer. Do NOT use simple tables. You MUST use the rich "Tier 1" layout from `_shared/output_conventions.md` §5.
+
+Format required:
+```markdown
+---
+## 📂 Generated Files (__YYYYMMDD__)
+> Auto-appended per `_shared/output_conventions.md` §5. Same-level snapshot of top-level command outputs at write time.
+
+### 📚 Base Reference Files
+| Type | File | Description |
+|---|---|---|
+| Foundational | [context.md](../../../../../context.md) | Permanent Identity and Architecture (Source of Truth) |
+| Snapshot | [context_<scope>_<date>.md](../contexts/context_<scope>_<date>.md) | Daily snapshot used for current session context |
+| Foundational | [dev.md](../../../../../dev.md) | Permanent Development Commands and Status |
+
+### Global Files (`core2/` monorepo root)
+| Category | File | Source Command |
+|---|---|---|
+| Reports | [audit_core2_<date>.md](../../../../../../../../../../.wb/workflows/reports/<YYYY>/<MM>/<DD>/audits/audit_core2_<date>.md) | `/wbAudit core2/` |
+| Reports | [plan_core2_<date>.md](../../../../../../../../../../.wb/workflows/reports/<YYYY>/<MM>/<DD>/plans/plan_core2_<date>.md) | `/wbPlan core2/` |
+| Tracks | [track_core2_<date>.md](../../../../../../../../../../.wb/workflows/tracks/<YYYY>/<MM>/<DD>/track_core2_<date>.md) | `/wbTrack core2/` |
+
+<details>
+  <summary>📂 Sub-Package: [Active Package Name]</summary>
+
+| Category | File | Source Command |
+|---|---|---|
+| Reports | [audit_subpkg_<date>.md](../../../../../../../../../../apps/wb-core/subpkg/.wb/workflows/reports/<YYYY>/<MM>/<DD>/audits/audit_subpkg_<date>.md) | `/wbAudit` |
+
+</details>
+```
+```

package/templates/commands/wbClean/wbClean_template.md

@@ -0,0 +1,118 @@
+# /wbClean: Execution Template
+
+
+<!-- HELP_GATE_START -->
+## Help intercept (handle FIRST — before any other action)
+
+**If `$ARGUMENTS` contains `--help`, `-h`, or `--h`** (case-insensitive, anywhere in the args), DO NOT execute the command's normal procedure. Instead:
+
+1. Output the **HELP BLOCK** below verbatim (rendered as markdown).
+2. Stop. Do not perform any file reads, writes, or other tool calls.
+3. Do not generate any reports under `.wb/workflows/reports/`.
+
+Otherwise, ignore this section and proceed to the rest of the template.
+
+### HELP BLOCK — `/wbClean`
+
+## One form
+
+```
+/wbClean <pkg-or-app>
+```
+
+No flags. The command is narrow: scan, report, done.
+
+## When to run
+
+- End of day / end of a work session — catches what you left behind while still fresh.
+- Before `/wbAudit` in the afternoon polish phase — clean first, then audit the clean result.
+- Before `/wbRelease` — confirm no `console.log` or debug code leaking to npm.
+- After inheriting a package — baseline the debt.
+
+## When *not* to run
+
+- Before starting work — nothing to clean.
+- On a package you rarely touch — low signal, wastes a report slot in `reports/`.
+- As a substitute for `/wbAudit` — clean finds debris; audit finds design flaws. Different problems.
+
+## Reading the output
+
+Five report sections:
+
+1. **Forgotten dev artifacts** — `console.log`, `debugger`, `TODO: hack` comments. Almost always safe to delete. HIGH confidence.
+2. **Dead files** — 0 import references detected. MEDIUM confidence (could be dynamic).
+3. **Unused imports** — imported, never referenced. HIGH confidence.
+4. **Commented-out blocks** — still there from weeks ago. MEDIUM confidence — sometimes intentional.
+5. **TODOs** — informational. Not a delete candidate; just a visibility surface.
+
+Also mandatory: **"What this clean did NOT check"** section declaring coverage gaps (dynamic refs, build-time includes, historical context).
+
+## The deletion step
+
+`/wbClean` doesn't delete. To actually remove things, follow up with an explicit instruction in the same session:
+
+```
+"Delete all HIGH-confidence items from the last clean report."
+```
+
+The two-step discipline (detect, then remove) prevents silent loss of code the AI misclassified.
+
+## The one mistake to avoid
+
+**Auto-deleting everything the report flags.** `/wbClean` can be wrong about dead files (dynamic imports, reflective code, build-time string references). HIGH-confidence items are nearly always safe; MEDIUM items need a human glance. A 10-second sanity check per MEDIUM item is cheap.
+
+## When /wbClean is the wrong command
+
+- Restructuring code without changing behavior → `/wbRefactor`.
+- Finding a specific bug → `/wbDebug`.
+- Verifying release-readiness → `/wbAudit`.
+- Checking tests pass → `/wbTest`.
+
+`/wbClean` answers one question: *"what debris should I delete?"*
+
+> For deeper reading: [`docs_claude/commands/wbClean/wbClean_practical_claude.md`](../../docs/docs_claude/commands/wbClean/wbClean_practical_claude.md) (or the `_eli5_`, `_expert_`, `_examples_` siblings).
+## Self-correct mode (dual-mode invocation)
+
+```
+/wbClean <scope_folder>           # normal mode — produce a fresh output file
+/wbClean <previous_output_file>   # self-correct mode — verify & repair the file in place
+```
+
+When the first arg is an existing output file from a prior `/wbClean` run (detected by its first H1 — see this template's **Detection** section), the command runs in **verify-and-repair** mode: gap-fills missing fields, normalizes links, ticks done/valid checkboxes whose reports exist, never rewrites authored content. See [`../_shared/output_conventions.md`](../_shared/output_conventions.md) §3.
+
+<!-- HELP_GATE_END -->
+
+**ROLE:** The Entropy Auditor
+**TARGET:** The provided directory path.
+**Read first:** [`../_shared/output_conventions.md`](../_shared/output_conventions.md) — applies to the report (relative links, full-syntax commands, self-correct mode).
+
+## ━━━ DETECTION (Self-Correct Mode) ━━━
+
+Trigger self-correct when the input file's first H1 matches:
+`# Clean: <scope> — <YYYY-MM-DD>` *(or the legacy `# Clean Entry #N` header).*
+
+Behavior is defined in [`../_shared/output_conventions.md`](../_shared/output_conventions.md) §3.
+
+Clean-specific gap-fills:
+
+- Plain-text candidate file paths → relative markdown links per §1.
+- Bare `/wbPlan` reference at the bottom → full-syntax `/wbPlan <target>` per §2.
+
+## ━━━ OBJECTIVE ━━━
+Your job is to scan the target directory for "Entropy" (dead code, unused files, empty folders, and obsolete dependencies) and generate a cleanup report. **DO NOT DELETE ANY FILES YOURSELF.** 
+
+## ━━━ PHASE 1: RECONNAISSANCE ━━━
+1. Read the `.wb/workflows/context.md` to understand what is actually important.
+2. Search for:
+   - Components or functions that are never imported.
+   - Files with `.bak`, `.old`, or commented-out massive blocks of code.
+   - Folders that are entirely empty or only contain useless wrappers.
+
+## ━━━ PHASE 2: REPORT GENERATION ━━━
+Create a report detailing your findings. 
+- **Path:** `.wb/workflows/reports/<YYYY>/<MM>/<DD>/cleans/clean_<target>_<YYYYMMDD>.md`
+- **No `<model>/` subfolder.** Create-or-append: if the file exists, append your clean report as the next Entry #N tagged `*(ModelName — HH:MM)*`.
+- **Smart Merge:** When appending, READ existing entries first. If you found the same dead-code candidate as a previous model (same file/function), add your vote to a **Model Votes** section and update the **Consensus Table** (`# | Candidate | Confidence | Models | Priority`) instead of duplicating. New unique findings get added as new rows. Add a merge log at the bottom.
+- **First model?** Just write normally as Entry #1. Consensus Table will be created by the second model.
+- **Format:** Group findings into categories: [High Priority (Breaking Debt)], [Medium (Clutter)], [Low (Cosmetic)]. Apply output_conventions.md §1 (relative links for every file path) and §2 (full-syntax for any /wb* command cited).
+- **Next Steps:** End the report with a `## 🧭 What's Next?` section: *"Run `/wbPlan <target>` to generate a safe execution matrix for these deletions, then `/wbNext <target>` for the broader picture."*

package/templates/commands/wbContext/wbContext_template.md

@@ -0,0 +1,213 @@
+# wbContext Template v2.1
+
+
+<!-- HELP_GATE_START -->
+## Help intercept (handle FIRST — before any other action)
+
+**If `$ARGUMENTS` contains `--help`, `-h`, or `--h`** (case-insensitive, anywhere in the args), DO NOT execute the command's normal procedure. Instead:
+
+1. Output the **HELP BLOCK** below verbatim (rendered as markdown).
+2. Stop. Do not perform any file reads, writes, or other tool calls.
+3. Do not generate any reports under `.wb/workflows/reports/`.
+
+Otherwise, ignore this section and proceed to the rest of the template.
+
+### HELP BLOCK — `/wbContext`
+
+## Three invocation forms
+
+```
+/wbContext <path>                          # standard — per-package
+/wbContext <path> --focus="<subsystem>"    # + focused sidecar file
+/wbContext core2/ --scope=global           # monorepo-wide survey
+```
+
+## When to run each
+
+| Situation | Form |
+|---|---|
+| Start of a fresh AI session | `/wbContext <current-pkg>` |
+| After pulling changes from git (if applicable) | `/wbContext <touched-pkgs>` |
+| Before touching a complex subsystem | `/wbContext <pkg> --focus="<subsystem>"` |
+| Monthly sanity check across the monorepo | `/wbContext core2/ --scope=global` |
+| You rewrote a lot of a package | `/wbContext <pkg>` (re-run, not `/wbSetup`) |
+
+## What you'll see in the output
+
+1. **Baseline load** — age of stored context.md / dev.md. If > 30 days and the package is active, flag it.
+2. **Drift report** — what changed between stored understanding and current code.
+3. **Questions** — if drift requires a decision (e.g., "did you rename this on purpose?"), AI asks before updating.
+4. **Report ingestion** — recent audits / debug reports surfaced into working memory.
+
+## The one mistake to avoid
+
+**Not answering the drift questions.** If the AI asks "was the rename intentional?" and you ignore it, `context.md` stays out of date. Every future session starts with the same stale baseline until you answer. Answer once, save yourself later.
+
+## When /wbContext is *not* the right command
+
+- You want to create context for a brand-new package → `/wbSetup`, not `/wbContext`.
+- You want to know what everyone else has been doing → `/wbStandup`, not `/wbContext`.
+- You want to verify everything is release-ready → `/wbAudit`, not `/wbContext`.
+
+`/wbContext` answers one question only: *"Is my stored understanding of this package still correct?"* Use it for that. Don't overload it.
+
+> For deeper reading: [`docs_claude/commands/wbContext/wbContext_practical_claude.md`](../../docs/docs_claude/commands/wbContext/wbContext_practical_claude.md) (or the `_eli5_`, `_expert_`, `_examples_` siblings).
+
+<!-- FLAGS_TABLE_START -->
+## Flags & shortcuts
+
+Both forms are equivalent — pass either:
+
+| Long form | Shortcut |
+|---|---|
+| `--focus` | `-f` |
+| `--scope` | `-s` |
+
+`-h` / `--help` / `--h` (any command) prints this help block instead of executing.
+## Self-correct mode (dual-mode invocation)
+
+```
+/wbContext <scope_folder>           # normal mode — produce a fresh output file
+/wbContext <previous_output_file>   # self-correct mode — verify & repair the file in place
+```
+
+When the first arg is an existing output file from a prior `/wbContext` run (detected by its first H1 — see this template's **Detection** section), the command runs in **verify-and-repair** mode: gap-fills missing fields, normalizes links, ticks done/valid checkboxes whose reports exist, never rewrites authored content. See [`../_shared/output_conventions.md`](../_shared/output_conventions.md) §3.
+
+<!-- FLAGS_TABLE_END -->
+<!-- HELP_GATE_END -->
+
+<!-- FLAG_NORMALIZE_START -->
+## Flag normalization (apply BEFORE parsing args)
+
+Before processing `$ARGUMENTS`, normalize these short-form flags to their long equivalents:
+
+- `-f` → `--focus`
+- `-s` → `--scope`
+
+The rest of this template documents only the long forms; the substitution above is the only place short forms are mentioned.
+<!-- FLAG_NORMALIZE_END -->
+
+
+
+> **How to use**: This is the manual template for generating a context file if you are not using the automated `/wbSetup` workflow. Paste this to an AI to have it generate a context report.
+> **Read first:** [`../_shared/output_conventions.md`](../_shared/output_conventions.md) — applies to every cell of the output (relative links, full-syntax commands, self-correct mode).
+
+---
+
+## Detection (Self-Correct Mode)
+
+Trigger self-correct when the input file's first H1 matches:
+`# Context: <scope> — <YYYY-MM-DD>` *(or the legacy `# Context Entry #N` header for entry-N append mode).*
+
+Behavior is defined in [`../_shared/output_conventions.md`](../_shared/output_conventions.md) §3.
+
+Context-specific gap-fills:
+
+- Plain-text mentions of files (`package.json`, `index.js`, dependencies) → relative markdown links per §1.
+- Bare commands referenced in narrative (e.g. "run `/wbAudit`") → full-syntax form per §2.
+
+---
+
+## Filename & Folder Convention (v2 — Universal Daily File)
+
+**Path:** `<target_folder>/.wb/workflows/reports/<YYYY>/<MM>/<DD>/contexts/context_<target>_<YYYYMMDD>.md`
+
+> **No `<model>/` subfolder.** All models contribute to ONE context file per day per scope.
+> **No timestamp in filename.** Filename uses only the date.
+
+**Create-or-Append Rule:**
+- **File does NOT exist →** CREATE the file with a header and your context analysis as Entry #1.
+- **File ALREADY exists →** READ it, then APPEND your context as the next Entry #N. You SHOULD correct or enrich previous entries if you see inaccuracies.
+- **Every entry tagged:** `*(ModelName via Client — HH:MM)*`
+
+**Example (v2):**
+If Claude Opus runs `/wbContext` on `packages/wb-core` on April 22, 2026:
+```
+packages/wb-core/.wb/workflows/reports/20260422/contexts/context_wb-core_20260422.md
+```
+
+---
+
+## The Prompt (copy from here ↓)
+
+```
+━━━━━━━━━━━━━ /wbContext ━━━━━━━━━━━━━
+
+📁 TARGET: __TARGET_FOLDER_PATH__
+📅 DATE: __TODAY__
+🤖 MODEL: __YOUR_MODEL_NAME__
+🖥️ CLIENT: __YOUR_CLIENT__
+
+━━━ INSTRUCTIONS ━━━
+
+**[TEMPORAL MEMORY]**: Before executing your main task, you MUST scan the `.wb/workflows/reports/` directory of your target scope. Look for recent `audits/`, `reviews/`, or `tests/` reports. Use these past reports to understand recent failures, technical debt, or architectural decisions that affect your current execution.
+
+
+1. Scan the target folder's `package.json`, `index.js`, or any configuration files to understand its architecture and purpose.
+2. Generate a comprehensive Context Report containing:
+   - **Identity:** What this package is and its role in the monorepo.
+   - **Dependencies:** What other monorepo packages it relies on.
+   - **Constraints:** Specific rules, frameworks, and linters applied here.
+
+3. SAVE using the Universal Daily File pattern:
+
+   CHECK if this file exists:
+   `__TARGET_FOLDER_PATH__/.wb/workflows/reports/__YYYY__/__MM__/__DD__/contexts/context___FOLDER_NAME_____YYYYMMDD__.md`
+
+   - If it does NOT exist → CREATE the file with header + your context as Entry #1
+   - If it ALREADY exists → APPEND your context as the next Entry #N
+
+   Entry header format (use `---` only when appending, NEVER as the very first line of the file):
+   # Context Entry #N — *(__YOUR_MODEL_NAME__ via __YOUR_CLIENT__ — __CURRENT_TIME__)*
+   > **Model:** __YOUR_MODEL_NAME__
+   > **Client:** __YOUR_CLIENT__
+   > **Time:** __TODAY__ __CURRENT_TIME__
+   [... your full context analysis ...]
+
+4. If this is the active context, ensure it is also reflected in the main `.wb/workflows/context.md` file for daily use.
+
+5. APPLY OUTPUT CONVENTIONS (see ../_shared/output_conventions.md):
+   - All file/folder references in the report → relative markdown links from the output file's directory (§1).
+   - Any /wb* commands cited → full-syntax form (§2).
+
+6. END THE FILE WITH:
+
+   ## 🧭 What's Next?
+
+   Run `/wbNext <target_folder>` to get a current, ranked list of next actions for this scope.
+
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+━━━ AUTO-APPEND FOOTER ━━━
+
+At the VERY END of the file (after "What's Next?"), you MUST append the `## 📂 Generated Files (__YYYYMMDD__)` cross-link footer. Do NOT use simple tables. You MUST use the rich "Tier 1" layout from `_shared/output_conventions.md` §5.
+
+Format required:
+```markdown
+---
+## 📂 Generated Files (__YYYYMMDD__)
+> Auto-appended per `_shared/output_conventions.md` §5. Same-level snapshot of top-level command outputs at write time.
+
+### 📚 Base Reference Files
+| Type | File | Description |
+|---|---|---|
+| Foundational | [context.md](../../../../../context.md) | Permanent Identity and Architecture (Source of Truth) |
+| Snapshot | [context_<scope>_<date>.md](../contexts/context_<scope>_<date>.md) | Daily snapshot used for current session context |
+| Foundational | [dev.md](../../../../../dev.md) | Permanent Development Commands and Status |
+
+### Global Files (`core2/` monorepo root)
+| Category | File | Source Command |
+|---|---|---|
+| Reports | [audit_core2_<date>.md](../../../../../../../../../../.wb/workflows/reports/<YYYY>/<MM>/<DD>/audits/audit_core2_<date>.md) | `/wbAudit core2/` |
+| Reports | [plan_core2_<date>.md](../../../../../../../../../../.wb/workflows/reports/<YYYY>/<MM>/<DD>/plans/plan_core2_<date>.md) | `/wbPlan core2/` |
+| Tracks | [track_core2_<date>.md](../../../../../../../../../../.wb/workflows/tracks/<YYYY>/<MM>/<DD>/track_core2_<date>.md) | `/wbTrack core2/` |
+
+<details>
+  <summary>📂 Sub-Package: [Active Package Name]</summary>
+
+| Category | File | Source Command |
+|---|---|---|
+| Reports | [audit_subpkg_<date>.md](../../../../../../../../../../apps/wb-core/subpkg/.wb/workflows/reports/<YYYY>/<MM>/<DD>/audits/audit_subpkg_<date>.md) | `/wbAudit` |
+
+</details>
+```
+```