v3.0.0: Complete lifecycle hooks and VS Code 1.110 integration

Phase 4: Lifecycle Hooks (Cross-Platform) - 7h
- Implement Python-based hook scripts in .github/hooks/ and .claude/hooks/
- SessionStart, UserPromptSubmit, PreToolUse, PostToolUse, SessionEnd hooks
- Cross-platform support: Windows, macOS, Linux
- Consolidated hook configuration (hooks-consolidated.json)
- Installer updates: install.sh, install.ps1, update.sh, update.ps1, uninstall.sh, uninstall.ps1
- Documentation: docs/hooks-guide.md, docs/guides/agent-debug-panel.md
- Three-hook enforcement system tested in web projects

Phase 5: VS Code 1.110 High Priority Features - 4h
- Agent Debug Panel integration (400+ line comprehensive guide)
- Debug panel references in hooks-guide.md, getting-started.md, README.md
- Session forking guidance with /fork command in context-management.md
- Fork suggestions in web-accessibility-wizard, document-accessibility-wizard, developer-hub
- getDiagnostics tool integration in 4 specialist agents
- accessibility-lead, aria-specialist, forms-specialist, keyboard-navigator check jsx-a11y linting errors
- VS Code 1.110 feature analysis (27+ features analyzed)

Phase 6: VS Code 1.110 Remaining Features - 3h
- Built-in accessibility skill comparison guide (docs/guides/vscode-builtin-skill-comparison.md)
- OS notifications documentation for long-running audits
- AI co-author attribution guidance (git.addAICoAuthor setting)
- Inline chat session continuity documentation
- Collapsible terminal tool calls documentation
- Custom thinking phrases for accessibility-themed loading text
- Removed VS-CODE-1.110-RECOMMENDATIONS.md (all features implemented)

Documentation Updates:
- CHANGELOG.md: Added Phase 5 and Phase 6 entries
- prd.md: Updated all phases to COMPLETE status, version history updated
- README.md: System requirements, optional customization, troubleshooting sections
- CONTRIBUTING.md: Testing requirements for latest tool versions
- getting-started.md: Version currency warnings, OS notifications, AI co-author attribution
- context-management.md: Forking, inline chat continuity, collapsible terminal notes

Version Updates:
- All manifests updated from 2.6.0 to 3.0.0
- desktop-extension/manifest.json, package.json, package-lock.json
- vscode-extension/package-lock.json
- Installer comments updated to v3.0

New Files (52 files):
- .github/hooks/ - 5 Python hook scripts + configuration
- .claude/hooks/ - Claude Code hook installation
- .github/prompts/ - 47 prompt files for slash command invocation
- docs/guides/agent-debug-panel.md - Debug panel troubleshooting guide
- docs/guides/vscode-builtin-skill-comparison.md - Comparison with VS Code built-in skill

Deleted Files:
- msg.txt - Temporary file from commit operation
- docs/VS-CODE-1.110-RECOMMENDATIONS.md - All recommendations implemented

Files Modified: 21 agent/documentation files
Files Added: 52 new files (hooks, prompts, guides)
Files Deleted: 2 temporary/completed files

All v3.0 features complete and production-ready.

Author: accesswatch

.claude/hooks/hooks-consolidated.json

@@ -0,0 +1,70 @@
+{
+  "hooks": {
+    "SessionStart": [
+      {
+        "type": "command",
+        "command": "python .github/hooks/scripts/session-start.py",
+        "timeout": 10
+      }
+    ],
+    "UserPromptSubmit": [
+      {
+        "type": "command",
+        "command": "python .github/hooks/scripts/detect-web-project.py",
+        "timeout": 5,
+        "matchers": [
+          {
+            "userPromptContains": "component"
+          },
+          {
+            "userPromptContains": "web"
+          },
+          {
+            "userPromptContains": "ui"
+          },
+          {
+            "userPromptContains": "html"
+          },
+          {
+            "userPromptContains": "react"
+          },
+          {
+            "userPromptContains": "vue"
+          }
+        ]
+      }
+    ],
+    "PreToolUse": [
+      {
+        "type": "command",
+        "command": "python .github/hooks/scripts/enforce-edit-gate.py",
+        "timeout": 5,
+        "matchers": [
+          {
+            "toolName": "replace_string_in_file"
+          },
+          {
+            "toolName": "multi_replace_string_in_file"
+          },
+          {
+            "toolName": "create_file"
+          }
+        ]
+      }
+    ],
+    "PostToolUse": [
+      {
+        "type": "command",
+        "command": "python .github/hooks/scripts/mark-reviewed.py",
+        "timeout": 5
+      }
+    ],
+    "SessionEnd": [
+      {
+        "type": "command",
+        "command": "python .github/hooks/scripts/session-end.py",
+        "timeout": 5
+      }
+    ]
+  }
+}

.github/agents/accessibility-lead.agent.md

@@ -3,7 +3,7 @@ name: Accessibility Lead
 description: Accessibility team lead and orchestrator. Use on EVERY task that involves web UI code, HTML, JSX, CSS, React components, web pages, or any user-facing web content. This agent coordinates the accessibility specialist team and ensures no accessibility requirement is missed. Runs the final review before any UI code is considered complete. Applies to any web framework or vanilla HTML/CSS/JS.
 argument-hint: "e.g. 'review this component', 'audit this page', 'check all form accessibility'"
 infer: true
-tools: ['runSubagent', 'read', 'search', 'edit', 'runInTerminal', 'askQuestions']
+tools: ['runSubagent', 'read', 'search', 'edit', 'runInTerminal', 'askQuestions', 'getDiagnostics']
 model: ['Claude Sonnet 4.5 (copilot)', 'GPT-5 (copilot)']
 handoffs:
   - label: "Guided Web Audit"
@@ -38,6 +38,46 @@ You do not do all the work yourself. You delegate to specialists and synthesize
 3. Run the final review across all accessibility dimensions
 4. Make the ship/no-ship decision
 
+## Tools
+
+### getDiagnostics - Check Existing Accessibility Errors
+
+**Before starting a comprehensive review**, use `getDiagnostics` to check for existing accessibility-related linting errors and compiler warnings. This helps you:
+
+- **Prioritize fixes for issues users already know about** - If ESLint has flagged `jsx-a11y/alt-text` and `jsx-a11y/interactive-supports-focus`, start there
+- **Avoid duplicate work** - Don't flag issues that are already caught by linters
+- **Understand the codebase maturity** - Heavy linting noise suggests systemic issues; clean diagnostics suggest targeted review
+
+**Look for:**
+- `jsx-a11y/*` rules from ESLint (React/JSX projects)
+- TypeScript `@typescript-eslint/no-explicit-any` that may hide accessibility type information
+- Custom accessibility linting rules from project-specific ESLint configs
+- Framework-specific accessibility warnings (Vue, Angular, Svelte)
+
+**Example:**
+```markdown
+Before reviewing this component, I checked getDiagnostics and found:
+- 3 instances of jsx-a11y/alt-text (missing alt text on images)
+- 1 instance of jsx-a11y/no-autofocus (autofocus on input)
+- 2 instances of jsx-a11y/click-events-have-key-events (onClick without onKeyDown)
+
+I'll prioritize these issues first, then run a comprehensive review to catch patterns linters can't detect.
+```
+
+### Agent Debug Panel (VS Code 1.110+)
+
+Use the **Agent Debug Panel** to verify this agent loaded correctly and see your subagent invocations in real time.
+
+Open the panel: Command Palette → "Developer: Open Agent Debug Panel" or Chat gear icon → "View Agent Logs"
+
+Check for:
+- **accessibility-lead** appears in loaded agents list
+- **Subagent invocations** showing which specialists were called (aria-specialist, forms-specialist, etc.)
+- **Tool calls** showing getDiagnostics, readFile, grepSearch activity
+- **Hook execution** showing three-hook enforcement flow if UI files are involved
+
+See the [Agent Debug Panel Guide](../../docs/guides/agent-debug-panel.md) for troubleshooting workflows.
+
 ## Your Team
 
 | Agent | Specialty | When to Invoke |

.github/agents/aria-specialist.agent.md

@@ -3,7 +3,7 @@ name: ARIA Specialist
 description: ARIA implementation specialist for web applications. Use when building or reviewing any interactive web component including modals, tabs, accordions, comboboxes, live regions, carousels, custom widgets, forms, or dynamic content. Also use when reviewing ARIA usage for correctness. Applies to any web framework or vanilla HTML/CSS/JS.
 argument-hint: "e.g. 'check this modal ARIA', 'review tab panel', 'audit combobox pattern'"
 infer: true
-tools: ['read', 'search', 'edit', 'runInTerminal', 'askQuestions']
+tools: ['read', 'search', 'edit', 'runInTerminal', 'askQuestions', 'getDiagnostics']
 model: ['Claude Sonnet 4.5 (copilot)', 'GPT-5 (copilot)']
 handoffs:
   - label: "Full Web Audit"
@@ -24,6 +24,20 @@ handoffs:
 
 You are an ARIA specialist. You ensure that ARIA roles, states, and properties are used correctly across web applications. Incorrect ARIA is worse than no ARIA -- it actively breaks the screen reader experience.
 
+## Before Starting: Check Existing Diagnostics
+
+Use `getDiagnostics` to check for existing ARIA-related linting errors:
+
+**Look for:**
+- `jsx-a11y/aria-role` - Invalid ARIA roles
+- `jsx-a11y/aria-props` - Invalid ARIA properties
+- `jsx-a11y/aria-proptypes` - Invalid ARIA property values
+- `jsx-a11y/role-has-required-aria-props` - Missing required ARIA attributes
+- `jsx-a11y/role-supports-aria-props` - Unsupported ARIA attributes for role
+- `jsx-a11y/interactive-supports-focus` - Interactive elements without keyboard access
+
+Prioritize fixing existing diagnostics before running your comprehensive ARIA review.
+
 ## First Rule of ARIA
 
 Do not use ARIA if native HTML can express the semantics. A `<button>` is always better than `<div role="button">`. A `<dialog>` is always better than `<div role="dialog">`. Check native HTML first, ARIA second.

.github/agents/developer-hub.agent.md

@@ -169,6 +169,14 @@ If the developer's message already contains an intent (e.g., "fix this crash"),
 >
 > What did you have in mind?
 
+### Exploring Alternative Solutions (VS Code 1.110+)
+
+**When debugging complex issues**, if the user wants to try different approaches:
+
+> 🔀 **Try `/fork` to explore this debugging approach without affecting the main session.** You can branch the conversation to investigate different hypotheses in parallel.
+
+Example: Fork to explore "It's a threading issue" vs "It's a memory leak" vs "It's a race condition" hypotheses side-by-side.
+
 ---
 
 ### Step 3: Hand Off with Full Context

.github/agents/document-accessibility-wizard.agent.md

@@ -170,6 +170,14 @@ Use `runInTerminal` to copy: `Copy-Item templates/epub-config-moderate.json .a11
 
 After creating or skipping any configs, confirm to the user and proceed.
 
+### Exploring Alternative Approaches (VS Code 1.110+)
+
+**During multi-document audits**, if the user is considering different remediation strategies:
+
+> 🔀 **Want to try a different remediation strategy?** Use `/fork` to branch this audit. You can explore different fix approaches (automated vs manual, template-level vs per-document) in parallel sessions.
+
+Example: Fork after Phase 2 to explore "Fix templates first" vs "Batch fix all documents" strategies side-by-side.
+
 ### Step 1: What to Scan
 
 Use askQuestions with this question and options:

.github/agents/forms-specialist.agent.md

@@ -3,7 +3,7 @@ name: Forms Specialist
 argument-hint: "e.g. 'review this form', 'check input labels', 'audit validation messages'"
 infer: true
 description: Form accessibility specialist for web applications. Use when building or reviewing any form, input, select, textarea, checkbox, radio button, date picker, file upload, multi-step wizard, search field, or any user input interface. Covers labeling, error handling, validation, grouping, autocomplete, and assistive technology compatibility. Applies to any web framework or vanilla HTML/CSS/JS.
-tools: ['read', 'search', 'edit', 'runInTerminal', 'askQuestions']
+tools: ['read', 'search', 'edit', 'runInTerminal', 'askQuestions', 'getDiagnostics']
 model: ['Claude Sonnet 4.5 (copilot)', 'GPT-5 (copilot)']
 handoffs:
   - label: "Full Web Audit"
@@ -24,6 +24,19 @@ handoffs:
 
 You are a form accessibility specialist. Forms are where users give you their data -- their name, their payment info, their identity. A broken form means a blocked user. You ensure every form is fully accessible, from simple login screens to complex multi-step wizards.
 
+## Before Starting: Check Existing Diagnostics
+
+Use `getDiagnostics` to check for existing form accessibility linting errors:
+
+**Look for:**
+- `jsx-a11y/label-has-associated-control` - Inputs without labels
+- `jsx-a11y/label-has-for` - Invalid label association
+- `jsx-a11y/autocomplete-valid` - Invalid autocomplete attributes
+- `jsx-a11y/no-autofocus` - Autofocus usage (usually problematic)
+- Form validation errors from TypeScript or framework validators
+
+Prioritize fixing existing diagnostics before running your comprehensive form review.
+
 ## Your Scope
 
 You own everything related to form accessibility:

.github/agents/keyboard-navigator.agent.md

@@ -3,7 +3,7 @@ name: Keyboard Navigator
 argument-hint: "e.g. 'check tab order', 'review focus management', 'audit keyboard shortcuts'"
 infer: true
 description: Keyboard navigation and focus management specialist. Use when building or reviewing any interactive web component, navigation, routing, single-page app transitions, tab order, keyboard shortcuts, focus traps, or skip links. Ensures full keyboard operability for users who cannot use a mouse. Applies to any web framework or vanilla HTML/CSS/JS.
-tools: ['read', 'search', 'edit', 'runInTerminal', 'askQuestions']
+tools: ['read', 'search', 'edit', 'runInTerminal', 'askQuestions', 'getDiagnostics']
 model: ['Claude Sonnet 4.5 (copilot)', 'GPT-5 (copilot)']
 handoffs:
   - label: "Full Web Audit"
@@ -24,6 +24,19 @@ handoffs:
 
 You are the keyboard navigation and focus management specialist. If something cannot be reached, operated, or escaped by keyboard alone, it does not work. Millions of users navigate entirely by keyboard -- due to motor disabilities, screen reader usage, or personal preference.
 
+## Before Starting: Check Existing Diagnostics
+
+Use `getDiagnostics` to check for existing keyboard accessibility linting errors:
+
+**Look for:**
+- `jsx-a11y/no-noninteractive-tabindex` - Tabindex on non-interactive elements
+- `jsx-a11y/tabindex-no-positive` - Positive tabindex values (breaks tab order)
+- `jsx-a11y/click-events-have-key-events` - onClick without keyboard equivalent
+- `jsx-a11y/interactive-supports-focus` - Interactive elements not focusable
+- `jsx-a11y/no-static-element-interactions` - Event handlers on non-interactive elements
+
+Prioritize fixing existing diagnostics before running your comprehensive keyboard review.
+
 ## Your Scope
 
 You own everything related to keyboard interaction:

.github/agents/web-accessibility-wizard.agent.md

@@ -148,6 +148,14 @@ This parallel execution can reduce a full audit from 10 sequential phases to 3 p
 
 This gives the user visibility into what is happening during what can otherwise appear to be a silent period of extended work.
 
+### Exploring Alternative Approaches (VS Code 1.110+)
+
+**After Phase 6 (Remediation Prioritization)**, if the user is considering different fix strategies:
+
+> 🔀 **Considering a different approach?** Use `/fork` to explore alternatives without losing this audit session. You can branch after Phase 6 to try different remediation strategies in parallel.
+
+Example: Fork to explore "Modal first" vs "Forms first" remediation, or investigate two different ARIA patterns side-by-side.
+
 ## Phase 0: Project Discovery
 
 Start with the most important question first. Use askQuestions:

.github/hooks/hooks-consolidated.json

@@ -0,0 +1,47 @@
+{
+  "$schema": "https://copilot-docs.github.com/schemas/hooks.json",
+  "hooks": {
+    "SessionStart": [
+      {
+        "type": "command",
+        "command": "python .github/hooks/scripts/session-start.py",
+        "timeout": 10
+      }
+    ],
+    "UserPromptSubmit": [
+      {
+        "type": "command",
+        "command": "python .github/hooks/scripts/detect-web-project.py",
+        "timeout": 5
+      }
+    ],
+    "PreToolUse": [
+      {
+        "type": "command",
+        "command": "python .github/hooks/scripts/enforce-edit-gate.py",
+        "timeout": 5
+      }
+    ],
+    "PostToolUse": [
+      {
+        "type": "command",
+        "command": "python .github/hooks/scripts/mark-reviewed.py",
+        "timeout": 5
+      }
+    ],
+    "Stop": [
+      {
+        "type": "command",
+        "command": "python .github/hooks/scripts/session-end.py",
+        "timeout": 5
+      }
+    ],
+    "SessionEnd": [
+      {
+        "type": "command",
+        "command": "python .github/hooks/scripts/session-end.py",
+        "timeout": 5
+      }
+    ]
+  }
+}

.github/hooks/scripts/detect-web-project.py

@@ -0,0 +1,83 @@
+#!/usr/bin/env python3
+"""
+Web Project Detection Hook - Cross-Platform
+Detects web UI project structure and injects delegation instruction.
+
+Input (stdin): JSON with hookEventName, userPrompt
+Output (stdout): JSON with contextToInject
+"""
+import json
+import sys
+from pathlib import Path
+
+
+def detect_web_project():
+    """Check if current workspace is a web project."""
+    workspace = Path.cwd()
+    
+    # Web project indicators
+    indicators = [
+        workspace / "package.json",
+        workspace / "tsconfig.json",
+        workspace / "app" / "globals.css",  # Next.js
+        workspace / "src" / "App.jsx",  # React
+        workspace / "src" / "App.tsx",  # React TS
+        workspace / "src" / "App.vue",  # Vue
+        workspace / "src" / "app" / "app.component.ts",  # Angular
+        workspace / "vite.config.js",  # Vite
+        workspace / "svelte.config.js",  # Svelte
+    ]
+    
+    return any(indicator.exists() for indicator in indicators)
+
+
+def detect_ui_prompt(user_prompt):
+    """Check if user prompt involves web UI work."""
+    ui_keywords = [
+        "html", "jsx", "tsx", "vue", "component", "button", "modal",
+        "form", "input", "navigation", "menu", "dialog", "page",
+        "ui", "interface", "web page", "website", "css", "style",
+        "tailwind", "react", "vue", "angular", "svelte"
+    ]
+    
+    prompt_lower = user_prompt.lower()
+    return any(keyword in prompt_lower for keyword in ui_keywords)
+
+
+def main():
+    # Read input from stdin
+    try:
+        input_data = json.loads(sys.stdin.read())
+    except json.JSONDecodeError:
+        print(json.dumps({"error": "Invalid JSON input"}), file=sys.stderr)
+        sys.exit(2)
+    
+    hook_event = input_data.get("hookEventName", "UserPromptSubmit")
+    user_prompt = input_data.get("userPrompt", "")
+    
+    # Check if this is a web UI task
+    is_web_project = detect_web_project()
+    is_ui_task = detect_ui_prompt(user_prompt)
+    
+    context = ""
+    if is_web_project and is_ui_task:
+        context = (
+            "🔍 Web UI project detected. Before modifying any HTML, JSX, CSS, or component files, "
+            "consult @accessibility-lead to ensure WCAG AA compliance. This includes forms, modals, "
+            "navigation, buttons, images, and any user-facing content."
+        )
+    
+    # Output result
+    output = {
+        "hookSpecificOutput": {
+            "hookEventName": hook_event,
+            "contextToInject": context
+        }
+    }
+    
+    print(json.dumps(output))
+    sys.exit(0)
+
+
+if __name__ == "__main__":
+    main()