@stella-studio/ck 1.0.0

package/.claude/scripts/ck-help.py

@@ -0,0 +1,1088 @@
+#!/usr/bin/env python3
+"""
+ClaudeKit Help Command - All-in-one guide with dynamic command discovery.
+Scans .claude/commands/ directory to build catalog at runtime.
+
+Usage:
+    python ck-help.py                    # Overview with quick start
+    python ck-help.py fix                # Category guide with workflow
+    python ck-help.py plan:fast          # Command details
+    python ck-help.py debug login error  # Task recommendations
+    python ck-help.py auth               # Search (unknown word)
+"""
+
+import sys
+import re
+import io
+from pathlib import Path
+
+# Fix Windows console encoding for Unicode characters
+if sys.platform == 'win32':
+    sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding='utf-8', errors='replace')
+
+
+# Output type markers for LLM presentation guidance
+# Format: @CK_OUTPUT_TYPE:<type>
+# Types:
+#   - comprehensive-docs: Full documentation, show verbatim + add context
+#   - category-guide: Workflow guide, show full + explain workflow
+#   - command-details: Single command, show + offer to run
+#   - search-results: Search matches, show + offer alternatives
+#   - task-recommendations: Task-based suggestions, explain reasoning
+OUTPUT_TYPES = {
+    "comprehensive-docs": "Show FULL output verbatim, then ADD helpful context, examples, and real-world tips",
+    "category-guide": "Show complete workflow, then ENHANCE with practical usage scenarios",
+    "command-details": "Show command info, then ADD usage examples and related commands",
+    "search-results": "Show all matches, then HELP user narrow down or explore",
+    "task-recommendations": "Show recommendations, then EXPLAIN why these fit and offer to start",
+}
+
+
+def emit_output_type(output_type: str) -> None:
+    """Emit output type marker for LLM presentation guidance."""
+    print(f"@CK_OUTPUT_TYPE:{output_type}")
+    print()
+
+
+# Fuzzy matching for typo tolerance
+def levenshtein_distance(s1: str, s2: str) -> int:
+    """Standard Levenshtein distance algorithm."""
+    if len(s1) < len(s2):
+        return levenshtein_distance(s2, s1)
+    if len(s2) == 0:
+        return len(s1)
+
+    prev_row = range(len(s2) + 1)
+    for i, c1 in enumerate(s1):
+        curr_row = [i + 1]
+        for j, c2 in enumerate(s2):
+            insertions = prev_row[j + 1] + 1
+            deletions = curr_row[j] + 1
+            substitutions = prev_row[j] + (c1 != c2)
+            curr_row.append(min(insertions, deletions, substitutions))
+        prev_row = curr_row
+    return prev_row[-1]
+
+
+def fuzzy_match(word: str, target: str, threshold: int = 2) -> bool:
+    """Check if word matches target within edit distance threshold."""
+    if len(word) < 3:  # Skip very short words
+        return word == target
+
+    # Exact match short-circuit
+    if word == target:
+        return True
+
+    # Require similar lengths to avoid false positives (e.g., "create" ≠ "creative")
+    len_diff = abs(len(word) - len(target))
+    if len_diff > 1:
+        return False
+
+    # Allow threshold based on word length (max 1/3 of target length)
+    max_edits = min(threshold, len(target) // 3)
+    if max_edits < 1:
+        return word == target
+
+    return levenshtein_distance(word, target) <= max_edits
+
+
+# Disambiguation threshold - if top 2 scores within this, ask user
+DISAMBIGUATION_THRESHOLD = 0.5
+
+
+# Synonym mappings for normalization (term → canonical)
+SYNONYMS = {
+    # Notifications
+    "alerts": "notifications",
+    "alert": "notification",
+
+    # Git/GitHub
+    "ci": "github actions",
+    "ci/cd": "github actions",
+    "pipeline": "github actions",
+    "actions": "github actions",
+
+    # Common abbreviations
+    "auth": "authentication",
+    "authn": "authentication",
+    "db": "database",
+    "repo": "repository",
+    "deps": "dependencies",
+
+    # Commands
+    "pr": "pull request",
+    "mr": "pull request",  # GitLab users
+
+    # Testing
+    "specs": "tests",
+    "e2e": "integration test",
+}
+
+
+def expand_synonyms(text: str) -> str:
+    """Replace synonyms with canonical terms."""
+    result = text.lower()
+
+    # Sort by length (longest first) to handle multi-word synonyms
+    sorted_synonyms = sorted(SYNONYMS.items(), key=lambda x: -len(x[0]))
+
+    for synonym, canonical in sorted_synonyms:
+        # Word boundary aware replacement
+        pattern = r'\b' + re.escape(synonym) + r'\b'
+        result = re.sub(pattern, canonical, result, flags=re.IGNORECASE)
+
+    return result
+
+
+# Task keyword mappings for intent detection
+TASK_MAPPINGS = {
+    "plan": ["plan", "design", "architect", "research", "think", "analyze", "strategy", "how to", "approach"],
+    "cook": ["implement", "build", "create", "add", "feature", "code", "develop", "make", "write"],
+    "bootstrap": ["start", "new", "init", "setup", "project", "scaffold", "generate", "begin"],
+    "test": ["test", "check", "verify", "validate", "spec", "unit", "integration", "coverage", "login", "auth", "e2e"],
+    "docs": ["document", "readme", "docs", "explain", "comment", "documentation"],
+    "review": ["review", "audit", "inspect", "quality", "refactor", "clean"],
+    "config": ["config", "configure", "settings", "ck.json", ".ck.json", "setup", "locale", "language", "paths"],
+    "coding-level": ["coding", "level", "eli5", "junior", "senior", "lead", "god", "beginner", "expert", "teach", "learn", "explain"],
+    # New categories
+    "worktree": ["worktree", "parallel", "isolate", "isolation", "concurrent", "multiple branches"],
+    "kanban": ["kanban", "board", "dashboard", "progress", "track", "orchestration", "visualize"],
+    "preview": ["preview", "view", "render", "markdown", "reader", "novel", "explain", "slides", "diagram", "ascii", "visualize", "visual"],
+    "journal": ["journal", "diary", "log", "entry", "reflect", "failure", "lesson"],
+    "watzup": ["watzup", "status", "summary", "wrap up", "what's up", "recent", "changes"],
+    "notifications": ["notification", "notifications", "notify", "discord", "telegram", "slack", "alert", "webhook", "stop hook", "session end", "setup notification", "setup notifications", "configure discord", "configure telegram", "configure slack", "discord webhook", "telegram bot", "slack webhook"],
+}
+
+# Category workflows and tips
+CATEGORY_GUIDES = {
+    "plan": {
+        "title": "Planning",
+        "workflow": [
+            ("Quick plan", "`/plan:fast` \"your task\""),
+            ("Deep research", "`/plan:hard` \"complex task\""),
+            ("Validate", "`/plan:validate` (interview to confirm decisions)"),
+            ("Execute plan", "`/cook` (runs the plan)"),
+        ],
+        "tip": "Use /plan:validate to confirm assumptions before coding",
+    },
+    "cook": {
+        "title": "Implementation",
+        "workflow": [
+            ("Quick impl", "`/cook` \"your feature\""),
+            ("Auto mode", "`/cook:auto` \"trust me bro\""),
+            ("Test", "`/test`"),
+        ],
+        "tip": "Cook is standalone - it plans internally. Use /plan → /cook for explicit planning",
+    },
+    "bootstrap": {
+        "title": "Project Setup",
+        "workflow": [
+            ("Quick start", "`/bootstrap:auto:fast` \"requirements\""),
+            ("Full setup", "`/bootstrap` \"detailed requirements\""),
+        ],
+        "tip": "Include tech stack preferences in description",
+    },
+    "test": {
+        "title": "Testing",
+        "workflow": [
+            ("Run tests", "`/test`"),
+        ],
+        "tip": "Run tests frequently during development",
+    },
+    "docs": {
+        "title": "Documentation",
+        "workflow": [
+            ("Initialize", "`/docs:init`"),
+            ("Update", "`/docs:update`"),
+        ],
+        "tip": "Keep docs close to code for accuracy",
+    },
+    "review": {
+        "title": "Code Review",
+        "workflow": [
+            ("Full review", "`/review:codebase`"),
+        ],
+        "tip": "Review before merging to main",
+    },
+    "coding-level": {
+        "title": "Coding Level (Adaptive Communication)",
+        "workflow": [
+            ("Disabled", "`codingLevel: -1` (default, no injection)"),
+            ("ELI5", "`codingLevel: 0` (analogies, baby steps)"),
+            ("Junior", "`codingLevel: 1` (WHY before HOW)"),
+            ("Mid-Level", "`codingLevel: 2` (patterns, trade-offs)"),
+            ("Senior", "`codingLevel: 3` (production code, ops)"),
+            ("Tech Lead", "`codingLevel: 4` (risk matrix, strategy)"),
+            ("God Mode", "`codingLevel: 5` (code first, no fluff)"),
+        ],
+        "tip": "Set in .ck.json. Guidelines auto-inject on session start",
+    },
+    "config": {
+        "title": "ClaudeKit Configuration (.ck.json)",
+        "workflow": [
+            ("Global", "Set user prefs in `~/.claude/.ck.json`"),
+            ("Local", "Override per-project in `./.claude/.ck.json`"),
+            ("Resolution", "DEFAULT → global → local (deep merge)"),
+        ],
+        "tip": "Global config works in fresh dirs; local overrides for projects",
+    },
+    # New category guides (workflow-first approach)
+    "worktree": {
+        "title": "Git Worktrees (Parallel Development)",
+        "workflow": [
+            ("Create worktree", "`/worktree` \"feature description\""),
+            ("Work in isolation", "cd to worktree, implement, test"),
+            ("Review & merge", "Create PR from worktree → merge → cleanup"),
+            ("List worktrees", "`/worktree list`"),
+            ("Remove worktree", "`/worktree remove <name>`"),
+        ],
+        "tip": "Use worktrees for parallel features without stashing. Each worktree = isolated branch + clean working directory",
+    },
+    "kanban": {
+        "title": "AI Orchestration Board",
+        "workflow": [
+            ("View dashboard", "`/kanban` (opens browser)"),
+            ("Specific plans", "`/kanban plans/my-feature/`"),
+            ("Track progress", "View phase completion, timeline, activity"),
+            ("Stop server", "`/kanban --stop`"),
+        ],
+        "tip": "Dashboard shows plan phases, progress bars, and agent activity. Future: worktree + agent orchestration",
+    },
+    "preview": {
+        "title": "Content Preview & Visual Generation",
+        "workflow": [
+            ("View markdown", "`/preview plans/plan.md`"),
+            ("Browse directory", "`/preview docs/`"),
+            ("Explain topic", "`/preview --explain OAuth flow`"),
+            ("Generate slides", "`/preview --slides API architecture`"),
+            ("Create diagram", "`/preview --diagram data flow`"),
+            ("ASCII only", "`/preview --ascii auth process`"),
+            ("Stop server", "`/preview --stop`"),
+        ],
+        "tip": "View existing markdown OR generate visual explanations (ASCII + Mermaid). Visuals save to active plan's visuals/ folder",
+    },
+    "journal": {
+        "title": "Technical Journaling",
+        "workflow": [
+            ("Write entry", "`/journal`"),
+            ("Document failures", "Capture what went wrong with emotional honesty"),
+            ("Lessons learned", "Turn setbacks into future guidance"),
+        ],
+        "tip": "Use after repeated test failures, critical bugs, or architectural pivots. Raw honesty = future wisdom",
+    },
+    "watzup": {
+        "title": "Session Review & Wrap-up",
+        "workflow": [
+            ("Review changes", "`/watzup`"),
+            ("Get summary", "See what was done, what files changed"),
+            ("Next steps", "Receive suggestions for what to do next"),
+        ],
+        "tip": "Run before ending session to capture progress and plan next steps",
+    },
+    "notifications": {
+        "title": "Session Notifications (Discord/Telegram/Slack)",
+        "workflow": [
+            ("1. Set env vars", "Add `DISCORD_WEBHOOK_URL` or `TELEGRAM_BOT_TOKEN`+`TELEGRAM_CHAT_ID` to `~/.claude/.env`"),
+            ("2. Add hook", "Add Stop hook to `.claude/settings.json` (see below)"),
+            ("3. Test", "`echo '{\"hook_event_name\":\"Stop\"}' | node .claude/hooks/notifications/notify.cjs`"),
+        ],
+        "tip": """Add to settings.json:
+```json
+"Stop": [{"matcher": "*", "hooks": [{"type": "command", "command": "node .claude/hooks/notifications/notify.cjs"}]}]
+```
+Docs: `.claude/hooks/notifications/docs/`""",
+    },
+}
+
+
+def detect_prefix(commands_dir: Path) -> str:
+    """Detect if commands use /ck: prefix based on directory structure."""
+    ck_commands_dir = commands_dir / "ck"
+    return "ck:" if ck_commands_dir.exists() and ck_commands_dir.is_dir() else ""
+
+
+def parse_frontmatter(file_path: Path) -> dict:
+    """Parse YAML frontmatter from a markdown file."""
+    try:
+        content = file_path.read_text(encoding='utf-8')
+    except Exception:
+        return {}
+
+    # Check for frontmatter
+    if not content.startswith('---'):
+        return {}
+
+    # Find closing ---
+    end_idx = content.find('---', 3)
+    if end_idx == -1:
+        return {}
+
+    frontmatter = content[3:end_idx].strip()
+    result = {}
+
+    for line in frontmatter.split('\n'):
+        if ':' in line:
+            key, value = line.split(':', 1)
+            result[key.strip()] = value.strip()
+
+    return result
+
+
+def discover_commands(commands_dir: Path, prefix: str) -> dict:
+    """Scan .claude/commands/ and build command catalog."""
+    commands = {}
+    categories = {}
+
+    if not commands_dir.exists():
+        return {"commands": commands, "categories": categories}
+
+    # Scan all .md files
+    for md_file in commands_dir.rglob("*.md"):
+        # Skip non-command files
+        rel_path = md_file.relative_to(commands_dir)
+        parts = rel_path.parts
+
+        # Get command name from path
+        # e.g., fix/fast.md -> fix:fast, plan.md -> plan
+        if len(parts) == 1:
+            # Root command: plan.md -> plan
+            cmd_name = parts[0].replace('.md', '')
+            category = "core"
+        else:
+            # Nested command: fix/fast.md -> fix:fast
+            category = parts[0]
+            cmd_name = ':'.join([p.replace('.md', '') for p in parts])
+
+        # Parse frontmatter
+        fm = parse_frontmatter(md_file)
+        description = fm.get('description', '')
+
+        # Skip if no description (not a real command)
+        if not description:
+            continue
+
+        # Clean description (remove emoji indicators)
+        clean_desc = re.sub(r'^[^\w\s]+\s*', '', description).strip()
+
+        # Format command name with prefix
+        formatted_name = f"/{prefix}{cmd_name}" if prefix else f"/{cmd_name}"
+
+        # Add to commands
+        if category not in commands:
+            commands[category] = []
+
+        commands[category].append({
+            "name": formatted_name,
+            "description": clean_desc,
+            "category": category,
+        })
+
+        # Track categories
+        if category not in categories:
+            categories[category] = category.title()
+
+    # Sort commands within each category
+    for cat in commands:
+        commands[cat].sort(key=lambda x: x["name"])
+
+    return {"commands": commands, "categories": categories}
+
+
+def detect_intent(input_str: str, categories: list) -> str:
+    """Smart auto-detection of user intent."""
+    if not input_str:
+        return "overview"
+
+    input_lower = input_str.lower()
+    words = input_str.split()
+
+    # Multiple words = likely task description (even if first word is a category)
+    # e.g., "test my login" should be task, not category "test"
+    if len(words) >= 2:
+        # Exception: if it looks like a command (has colon), treat as command
+        if ':' in input_str:
+            return "command"
+        return "task"
+
+    # Single word: check if it's a known category from discovered commands
+    if input_lower in [c.lower() for c in categories]:
+        return "category"
+
+    # Check if it's a known category from CATEGORY_GUIDES (includes non-command categories)
+    if input_lower in [c.lower() for c in CATEGORY_GUIDES.keys()]:
+        return "category"
+
+    # Single word typo tolerance: fuzzy match against categories
+    all_categories = set(c.lower() for c in categories) | set(c.lower() for c in CATEGORY_GUIDES.keys())
+    for cat in all_categories:
+        if fuzzy_match(input_lower, cat):
+            return "category"
+
+    # Single word typo tolerance: fuzzy match against task keywords
+    for cat, keywords in TASK_MAPPINGS.items():
+        for kw in keywords:
+            if ' ' not in kw and fuzzy_match(input_lower, kw):
+                return "task"
+
+    # Check if it looks like a command (has colon)
+    if ':' in input_str:
+        return "command"
+
+    return "search"
+
+
+def show_overview(data: dict, prefix: str) -> None:
+    """Display overview with quick start guide."""
+    emit_output_type("category-guide")
+
+    commands = data["commands"]
+    categories = data["categories"]
+    total = sum(len(cmds) for cmds in commands.values())
+    help_cmd = f"/{prefix}ck-help" if prefix else "/ck-help"
+
+    print("# ClaudeKit Commands")
+    print()
+    print(f"{total} commands across {len(categories)} categories.")
+    print()
+    print("**Quick Start:**")
+    print(f"- `/{prefix}cook` - Implement features (standalone)")
+    print(f"- `/{prefix}plan` + `/{prefix}cook` - Plan then execute")
+    print(f"- `/{prefix}test` - Run and analyze tests")
+    print()
+    print("**Common Workflows:**")
+    print(f"- New feature: `/{prefix}plan` → `/{prefix}cook` → `/{prefix}test`")
+    print(f"- Review: `/{prefix}review` → `/{prefix}watzup`")
+    print()
+    print("**Categories:**")
+    for cat_key in sorted(categories.keys()):
+        count = len(commands.get(cat_key, []))
+        print(f"- `{cat_key}` ({count})")
+    print()
+    print("**Usage:**")
+    print(f"- `{help_cmd} <category>` - Category guide with workflow")
+    print(f"- `{help_cmd} <command>` - Command details")
+    print(f"- `{help_cmd} <task description>` - Recommendations")
+    print()
+    print("**Tips:**")
+    print(f"- Unclear about approach? → `/{prefix}brainstorm` first")
+    print(f"- Agent generated report? → `/{prefix}preview` to view")
+    print("- Add `ultrathink` for deep analysis (more tokens)")
+    print("- `:parallel` variants (e.g., `/code:parallel`) = faster but more tokens, check quota")
+
+
+def show_category_guide(data: dict, category: str, prefix: str) -> None:
+    """Display category guide with workflow and tips."""
+    emit_output_type("category-guide")
+
+    categories = data["categories"]
+    commands = data["commands"]
+
+    # Find matching category (case-insensitive) - check both discovered and CATEGORY_GUIDES
+    cat_key = None
+    category_lower = category.lower()
+
+    for key in categories:
+        if key.lower() == category_lower:
+            cat_key = key
+            break
+
+    # Also check CATEGORY_GUIDES for categories without discovered commands (worktree, kanban, etc.)
+    if not cat_key:
+        for key in CATEGORY_GUIDES.keys():
+            if key.lower() == category_lower:
+                cat_key = key
+                break
+
+    # Fuzzy match for typos (e.g., "notifcations" → "notifications")
+    if not cat_key:
+        all_categories = list(categories.keys()) + list(CATEGORY_GUIDES.keys())
+        for key in all_categories:
+            if fuzzy_match(category_lower, key.lower()):
+                cat_key = key
+                break
+
+    if not cat_key:
+        all_cats = set(categories.keys()) | set(CATEGORY_GUIDES.keys())
+        print(f"Category '{category}' not found.")
+        print()
+        print("Available: " + ", ".join(f"`{c}`" for c in sorted(all_cats)))
+        return
+
+    cmds = commands.get(cat_key, [])
+    guide = CATEGORY_GUIDES.get(cat_key, {})
+
+    print(f"# {guide.get('title', cat_key.title())}")
+    print()
+
+    # Workflow first (most important)
+    if "workflow" in guide:
+        print("**Workflow:**")
+        for step, cmd in guide["workflow"]:
+            print(f"- {step}: {cmd}")
+        print()
+
+    # Commands list (only if we have discovered commands for this category)
+    if cmds:
+        print("**Commands:**")
+        for cmd in cmds:
+            print(f"- `{cmd['name']}` - {cmd['description']}")
+
+    # Tip at the end
+    if "tip" in guide:
+        print()
+        print(f"*Tip: {guide['tip']}*")
+
+
+def show_command(data: dict, command: str, prefix: str) -> None:
+    """Display command details."""
+    emit_output_type("command-details")
+
+    commands = data["commands"]
+
+    # Normalize search term
+    search = command.lower().replace("/ck:", "").replace("/", "").replace(":", "")
+
+    found = None
+    for cmds in commands.values():
+        for cmd in cmds:
+            # Normalize command name for comparison
+            name = cmd["name"].lower().replace("/ck:", "").replace("/", "").replace(":", "")
+            if name == search:
+                found = cmd
+                break
+        if found:
+            break
+
+    if not found:
+        print(f"Command '{command}' not found.")
+        print()
+        do_search(data, command.replace(":", " "), prefix)
+        return
+
+    print(f"# `{found['name']}`")
+    print()
+    print(found['description'])
+    print()
+    print(f"**Category:** {found['category']}")
+    print()
+    print(f"**Usage:** `{found['name']} <your-input>`")
+
+    # Show related commands (same category)
+    cat = found['category']
+    if cat in commands:
+        related = [c for c in commands[cat] if c['name'] != found['name']][:3]
+        if related:
+            related_names = ", ".join(f"`{r['name']}`" for r in related)
+            print()
+            print(f"**Related:** {related_names}")
+
+
+def do_search(data: dict, term: str, prefix: str) -> None:
+    """Search commands by keyword."""
+    emit_output_type("search-results")
+
+    commands = data["commands"]
+    term_lower = term.lower()
+    matches = []
+
+    for cmds in commands.values():
+        for cmd in cmds:
+            if term_lower in cmd["name"].lower() or term_lower in cmd["description"].lower():
+                matches.append(cmd)
+
+    if not matches:
+        print(f"No commands found for '{term}'.")
+        print()
+        print("Try browsing categories: " + ", ".join(f"`{c}`" for c in sorted(data["categories"].keys())))
+        return
+
+    print(f"# Search: {term}")
+    print()
+    print(f"Found {len(matches)} matches:")
+    for cmd in matches[:8]:
+        print(f"- `{cmd['name']}` - {cmd['description']}")
+
+
+def format_disambiguation(task: str, candidates: list) -> None:
+    """Output disambiguation prompt for close-scoring categories."""
+    emit_output_type("task-recommendations")
+
+    print(f"# Clarify: {task}")
+    print()
+    print("Your query matches multiple categories. Which did you mean?")
+    print()
+
+    for i, (cat, score) in enumerate(candidates[:3], 1):
+        guide = CATEGORY_GUIDES.get(cat, {})
+        title = guide.get("title", cat.title())
+        # Show first workflow step as example
+        example = ""
+        if "workflow" in guide and guide["workflow"]:
+            example = f" (e.g., {guide['workflow'][0][1]})"
+        print(f"{i}. **{title}**{example}")
+
+    print()
+    print("*Reply with the number or rephrase your question.*")
+
+
+def recommend_task(data: dict, task: str, prefix: str) -> None:
+    """Recommend commands for a task description."""
+    emit_output_type("task-recommendations")
+
+    commands = data["commands"]
+
+    # Expand synonyms first, then lowercase
+    task_expanded = expand_synonyms(task)
+    task_lower = task_expanded
+    words = task_lower.split()
+
+    # Action verbs that indicate primary intent when at sentence start
+    # These get BONUS weight when they appear first (imperative sentences)
+    # NOTE: Excluded contextual words like "setup", "add" that often precede subjects
+    ACTION_VERBS = {
+        "fix", "debug", "test", "commit", "push", "merge", "pull", "create",
+        "build", "implement", "write", "make", "deploy", "run",
+        "configure", "install", "update", "upgrade", "delete", "remove",
+        "review", "check", "verify", "validate", "find", "search", "locate",
+        "plan", "design", "refactor", "optimize", "document", "explain",
+    }
+
+    # Check if first word is an action verb
+    first_word_is_action = words[0] in ACTION_VERBS if words else False
+
+    # Score categories by keyword matches with smart weighting
+    scores = {}
+    for cat, keywords in TASK_MAPPINGS.items():
+        score = 0.0
+        for kw in keywords:
+            # Multi-word keywords: exact substring match, high weight
+            if ' ' in kw:
+                if kw in task_lower:
+                    score += 3.0
+            # Single-word keywords: exact match first, then fuzzy fallback
+            else:
+                matched_pos = -1
+                is_fuzzy = False
+
+                # Try exact match first
+                match = re.search(r'\b' + re.escape(kw) + r'\b', task_lower)
+                if match:
+                    # Find word position from character position
+                    char_count = 0
+                    for i, word in enumerate(words):
+                        if char_count <= match.start() < char_count + len(word):
+                            matched_pos = i
+                            break
+                        char_count += len(word) + 1
+                else:
+                    # Fuzzy matching fallback for typos
+                    for i, word in enumerate(words):
+                        if fuzzy_match(word, kw):
+                            matched_pos = i
+                            is_fuzzy = True
+                            break
+
+                if matched_pos >= 0:
+                    # Smart weighting based on sentence structure
+                    if len(words) > 1:
+                        if first_word_is_action and matched_pos == 0:
+                            weight = 2.5
+                        elif first_word_is_action:
+                            weight = 1.0
+                        else:
+                            weight = 1.0 + (matched_pos / (len(words) - 1))
+                    else:
+                        weight = 2.0
+
+                    # Slight penalty for fuzzy matches (0.8x)
+                    if is_fuzzy:
+                        weight *= 0.8
+
+                    score += weight
+        if score > 0:
+            scores[cat] = score
+
+    if not scores:
+        print(f"Not sure about: {task}")
+        print()
+        print("Try being more specific, or browse categories: " + ", ".join(f"`{c}`" for c in sorted(data["categories"].keys())))
+        return
+
+    sorted_cats = sorted(scores.items(), key=lambda x: -x[1])
+
+    # Check for ambiguity - if top 2 scores are close, ask user to clarify
+    if len(sorted_cats) >= 2:
+        top_score = sorted_cats[0][1]
+        second_score = sorted_cats[1][1]
+
+        # If scores too close, disambiguate
+        if top_score - second_score < DISAMBIGUATION_THRESHOLD and top_score > 0:
+            format_disambiguation(task, sorted_cats[:3])
+            return
+
+    top_cat = sorted_cats[0][0]
+    guide = CATEGORY_GUIDES.get(top_cat, {})
+
+    print(f"# Recommended for: {task}")
+    print()
+
+    # Show workflow first (most actionable)
+    if "workflow" in guide:
+        print("**Workflow:**")
+        for step, cmd in guide["workflow"][:3]:
+            print(f"- {step}: {cmd}")
+        print()
+
+    # Show relevant commands - only from top matched category
+    # Avoid showing unrelated commands from secondary matches
+    if top_cat in commands and commands[top_cat]:
+        print("**Commands:**")
+        for cmd in commands[top_cat][:4]:
+            print(f"- `{cmd['name']}` - {cmd['description']}")
+
+    if "tip" in guide:
+        print()
+        print(f"*Tip: {guide['tip']}*")
+
+
+def show_config_guide() -> None:
+    """Display comprehensive .ck.json configuration guide."""
+    emit_output_type("comprehensive-docs")
+
+    print("# ClaudeKit Configuration (.ck.json)")
+    print()
+    print("**Locations (cascading resolution):**")
+    print("- Global: `~/.claude/.ck.json` (user preferences)")
+    print("- Local: `./.claude/.ck.json` (project overrides)")
+    print()
+    print("**Resolution Order:** `DEFAULT → global → local`")
+    print("- Global config sets user defaults")
+    print("- Local config overrides for specific projects")
+    print("- Deep merge: nested objects merge recursively")
+    print()
+    print("**Purpose:** Customize plan naming, paths, locale, and hook behavior.")
+    print()
+    print("---")
+    print()
+    print("## Quick Start")
+    print()
+    print("**Global config** (`~/.claude/.ck.json`) - your preferences:")
+    print("```json")
+    print('{')
+    print('  "locale": {')
+    print('    "thinkingLanguage": "en",')
+    print('    "responseLanguage": "vi"')
+    print('  },')
+    print('  "plan": { "issuePrefix": "GH-" }')
+    print('}')
+    print("```")
+    print()
+    print("**Local override** (`./.claude/.ck.json`) - project-specific:")
+    print("```json")
+    print('{')
+    print('  "plan": { "issuePrefix": "JIRA-" },')
+    print('  "paths": { "docs": "documentation" }')
+    print('}')
+    print("```")
+    print()
+    print("---")
+    print()
+    print("## Full Schema")
+    print()
+    print("```json")
+    print('{')
+    print('  "plan": {')
+    print('    "namingFormat": "{date}-{issue}-{slug}",  // Plan folder naming')
+    print('    "dateFormat": "YYMMDD-HHmm",              // Date format in names')
+    print('    "issuePrefix": "GH-",                     // Issue ID prefix (null = #)')
+    print('    "reportsDir": "reports",                  // Reports subfolder')
+    print('    "resolution": {')
+    print('      "order": ["session", "branch"],  // Resolution chain')
+    print('      "branchPattern": "(?:feat|fix|...)/.+"  // Branch slug regex')
+    print('    },')
+    print('    "validation": {')
+    print('      "mode": "prompt",       // "auto" | "prompt" | "off"')
+    print('      "minQuestions": 3,      // Min questions to ask')
+    print('      "maxQuestions": 8,      // Max questions to ask')
+    print('      "focusAreas": ["assumptions", "risks", "tradeoffs", "architecture"]')
+    print('    }')
+    print('  },')
+    print('  "paths": {')
+    print('    "docs": "docs",     // Documentation directory')
+    print('    "plans": "plans"    // Plans directory')
+    print('  },')
+    print('  "locale": {')
+    print('    "thinkingLanguage": null, // Language for reasoning ("en" recommended)')
+    print('    "responseLanguage": null  // Language for output ("vi", "fr", etc.)')
+    print('  },')
+    print('  "trust": {')
+    print('    "passphrase": null,   // Secret for testing context injection')
+    print('    "enabled": false      // Enable trust verification')
+    print('  },')
+    print('  "project": {')
+    print('    "type": "auto",           // "monorepo", "single-repo", "auto"')
+    print('    "packageManager": "auto", // "npm", "pnpm", "yarn", "auto"')
+    print('    "framework": "auto"       // "next", "react", "vue", "auto"')
+    print('  },')
+    print('  "codingLevel": -1  // Adaptive communication (-1 to 5)')
+    print('}')
+    print("```")
+    print()
+    print("---")
+    print()
+    print("## Key Concepts")
+    print()
+    print("**Plan Resolution Chain:**")
+    print("1. `session` - Check session temp file for active plan")
+    print("2. `branch` - Match git branch slug to plan folder")
+    print()
+    print("**Naming Format Variables:**")
+    print("- `{date}` - Formatted date (per dateFormat)")
+    print("- `{issue}` - Issue ID with prefix")
+    print("- `{slug}` - Descriptive slug from branch or input")
+    print()
+    print("**Language Settings:**")
+    print("- `thinkingLanguage` - Language for internal reasoning (\"en\" recommended)")
+    print("- `responseLanguage` - Language for user-facing output (\"vi\", \"fr\", etc.)")
+    print()
+    print("When both are set, Claude thinks in one language but responds in another.")
+    print("This improves precision (English) while maintaining natural output (your language).")
+    print()
+    print("**Plan Validation:**")
+    print("- `mode: \"prompt\"` - Ask user after plan creation (default)")
+    print("- `mode: \"auto\"` - Always run validation interview")
+    print("- `mode: \"off\"` - Skip; user runs `/plan:validate` manually")
+    print()
+    print("Validation interviews the user with critical questions to confirm")
+    print("assumptions, risks, and architectural decisions before implementation.")
+    print()
+    print("**Coding Level (Adaptive Communication):**")
+    print("- `-1` = Disabled (default) - no injection, saves tokens")
+    print("- `0` = ELI5 - analogies, baby steps, check-ins")
+    print("- `1` = Junior - WHY before HOW, pitfalls, takeaways")
+    print("- `2` = Mid-Level - patterns, trade-offs, scalability")
+    print("- `3` = Senior - trade-offs table, production code, ops")
+    print("- `4` = Tech Lead - executive summary, risk matrix, business impact")
+    print("- `5` = God Mode - code first, minimal prose, no hand-holding")
+    print()
+    print("Guidelines auto-inject on session start. Commands like `/brainstorm` respect them.")
+    print()
+    print("---")
+    print()
+    print("## Edge Cases & Validation")
+    print()
+    print("**Path Handling:**")
+    print("- Trailing slashes normalized (`plans/` → `plans`)")
+    print("- Empty/whitespace-only paths fall back to defaults")
+    print("- Absolute paths supported (e.g., `/home/user/all-plans`)")
+    print("- Path traversal (`../`) blocked for relative paths")
+    print("- Null bytes and control chars rejected")
+    print()
+    print("**Slug Sanitization:**")
+    print("- Invalid filename chars removed: `< > : \" / \\ | ? *`")
+    print("- Non-alphanumeric replaced with hyphen")
+    print("- Multiple hyphens collapsed: `foo---bar` → `foo-bar`")
+    print("- Leading/trailing hyphens removed")
+    print("- Max 100 chars to prevent filesystem issues")
+    print()
+    print("**Naming Pattern Validation:**")
+    print("- Pattern must contain `{slug}` placeholder")
+    print("- Result must be non-empty after variable substitution")
+    print("- Unresolved placeholders (except `{slug}`) trigger error")
+    print("- Malformed JSON config falls back to defaults")
+    print()
+    print("**Consolidated Plans (advanced):**")
+    print("```json")
+    print('{')
+    print('  "paths": {')
+    print('    "plans": "/home/user/all-my-plans"')
+    print('  }')
+    print('}')
+    print("```")
+    print("Absolute paths allow storing all plans in one location across projects.")
+    print()
+    print("---")
+    print()
+    print("## Examples")
+    print()
+    print("**Global install user (fresh directories work):**")
+    print("```bash")
+    print("# ~/.claude/.ck.json - applies everywhere")
+    print("cd /tmp/new-project && claude  # Uses global config")
+    print("```")
+    print()
+    print("**Project with local override:**")
+    print("```bash")
+    print("# Global: issuePrefix = \"GH-\"")
+    print("# Local (.claude/.ck.json): issuePrefix = \"JIRA-\"")
+    print("# Result: issuePrefix = \"JIRA-\" (local wins)")
+    print("```")
+    print()
+    print("**Deep merge behavior:**")
+    print("```")
+    print("Global: { plan: { issuePrefix: \"GH-\", dateFormat: \"YYMMDD\" } }")
+    print("Local:  { plan: { issuePrefix: \"JIRA-\" } }")
+    print("Result: { plan: { issuePrefix: \"JIRA-\", dateFormat: \"YYMMDD\" } }")
+    print("```")
+    print()
+    print("*Tip: Config is optional - all fields have sensible defaults.*")
+
+
+def show_coding_level_guide() -> None:
+    """Display comprehensive codingLevel configuration guide."""
+    emit_output_type("comprehensive-docs")
+
+    print("# Coding Level (Adaptive Communication)")
+    print()
+    print("Adjusts Claude's communication style based on user's experience level.")
+    print("Guidelines auto-inject on SessionStart. Commands respect them.")
+    print()
+    print("---")
+    print()
+    print("## Levels")
+    print()
+    print("| Level | Name | Description |")
+    print("|-------|------|-------------|")
+    print("| **-1** | **Disabled** | Default - no injection, saves tokens |")
+    print("| 0 | ELI5 | Zero experience - analogies, baby steps, check-ins |")
+    print("| 1 | Junior | 0-2 years - WHY before HOW, pitfalls, takeaways |")
+    print("| 2 | Mid-Level | 3-5 years - patterns, trade-offs, scalability |")
+    print("| 3 | Senior | 5-8 years - trade-offs table, production code, ops |")
+    print("| 4 | Tech Lead | 8-15 years - executive summary, risk matrix, strategy |")
+    print("| 5 | God Mode | 15+ years - code first, minimal prose, no fluff |")
+    print()
+    print("---")
+    print()
+    print("## Configuration")
+    print()
+    print("**Set in `.ck.json`:**")
+    print("```json")
+    print('{')
+    print('  "codingLevel": 0')
+    print('}')
+    print("```")
+    print()
+    print("**Location (cascading):**")
+    print("- Global: `~/.claude/.ck.json` - personal preference")
+    print("- Local: `./.claude/.ck.json` - project override")
+    print()
+    print("---")
+    print()
+    print("## How It Works")
+    print()
+    print("1. SessionStart hook reads `codingLevel` from `.ck.json`")
+    print("2. If 0-5, injects guidelines from `.claude/output-styles/coding-level-*.md`")
+    print("3. Commands like `/brainstorm` follow the injected guidelines")
+    print()
+    print("**Token Efficiency:**")
+    print("- `-1` (default): Zero injection, zero overhead")
+    print("- `0-5`: Only selected level's guidelines injected (not all)")
+    print()
+    print("---")
+    print()
+    print("## Level Details")
+    print()
+    print("### Level 0 (ELI5)")
+    print("- **MUST** use real-world analogies (labeled boxes, recipes)")
+    print("- **MUST** define every technical term")
+    print("- **MUST** use \"we\" language")
+    print("- **MUST** end with check-in: \"Does this make sense?\"")
+    print("- **MUST** comment every line of code")
+    print("- Structure: Big Picture → Analogy → Baby Steps → Try It → Check-In")
+    print()
+    print("### Level 1 (Junior)")
+    print("- Explain WHY before HOW")
+    print("- Define technical terms on first use")
+    print("- Include common pitfalls section")
+    print("- Add Key Takeaways and Learn More links")
+    print("- Structure: Context → Approach → Implementation → Pitfalls → Takeaways")
+    print()
+    print("### Level 2 (Mid-Level)")
+    print("- Discuss design patterns and when to use them")
+    print("- Highlight trade-offs explicitly")
+    print("- Consider scalability implications")
+    print("- Reference patterns by name")
+    print("- Structure: Approach → Design → Implementation → Edge Cases")
+    print()
+    print("### Level 3 (Senior)")
+    print("- Lead with trade-offs table")
+    print("- Show production-ready code")
+    print("- Discuss operational concerns (monitoring, logging)")
+    print("- Flag security implications")
+    print("- **NEVER** explain basic concepts")
+    print("- Structure: Trade-offs → Implementation → Ops → Security")
+    print()
+    print("### Level 4 (Tech Lead)")
+    print("- Executive summary first (3-4 sentences)")
+    print("- Risk assessment matrix (Likelihood × Impact)")
+    print("- Strategic options comparison")
+    print("- Business impact analysis")
+    print("- Identify decisions needing stakeholder alignment")
+    print("- Structure: Summary → Risks → Options → Approach → Business Impact")
+    print()
+    print("### Level 5 (God Mode)")
+    print("- Answer exactly what was asked, nothing more")
+    print("- Code first, minimal prose")
+    print("- No explanations unless asked")
+    print("- **NEVER** use filler phrases")
+    print("- Trust their judgment completely")
+    print()
+    print("---")
+    print()
+    print("## Customization")
+    print()
+    print("Guidelines live in `.claude/output-styles/coding-level-*.md`")
+    print("Edit these files directly to customize behavior per level.")
+    print()
+    print("*Tip: Use `-1` (disabled) unless you're teaching or want guided explanations.*")
+
+
+def main():
+    # Find .claude/commands directory
+    script_path = Path(__file__).resolve()
+    claude_dir = script_path.parent.parent  # .claude/scripts -> .claude
+    commands_dir = claude_dir / "commands"
+
+    if not commands_dir.exists():
+        print("Error: .claude/commands/ directory not found.")
+        sys.exit(1)
+
+    # Detect prefix and discover commands
+    prefix = detect_prefix(commands_dir)
+    data = discover_commands(commands_dir, prefix)
+
+    if not data["commands"]:
+        print("No commands found in .claude/commands/")
+        sys.exit(1)
+
+    # Parse input
+    args = sys.argv[1:]
+    input_str = " ".join(args).strip()
+
+    # Special case: config documentation (not a command category)
+    if input_str.lower() in ["config", "configuration", ".ck.json", "ck.json"]:
+        show_config_guide()
+        return
+
+    # Special case: coding level documentation
+    if input_str.lower() in ["coding-level", "codinglevel", "coding level", "level", "eli5", "god mode"]:
+        show_coding_level_guide()
+        return
+
+    # Detect intent and route
+    intent = detect_intent(input_str, list(data["categories"].keys()))
+
+    if intent == "overview":
+        show_overview(data, prefix)
+    elif intent == "category":
+        show_category_guide(data, input_str, prefix)
+    elif intent == "command":
+        show_command(data, input_str, prefix)
+    elif intent == "task":
+        recommend_task(data, input_str, prefix)
+    else:
+        do_search(data, input_str, prefix)
+
+
+if __name__ == "__main__":
+    main()