zhongwei/gh-shakes-tzd-contextune
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
master
gh-shakes-tzd-contextune/commands/ctx-help.md
Zhongwei Li 400ca062d1 Initial commit
2025-11-30 08:56:10 +08:00

7.8 KiB
Raw Permalink Blame History

name, description, keywords
name description keywords
ctx:help Example-first command reference and quick start guide
help
examples
quick start
how to use
show examples
command reference
getting started

Contextune Help - Quick Start Guide

Just type naturally—Contextune detects your intent automatically!

✨ Try These Examples (Copy & Paste)

🔍 Fast Research (1-2 min, ~$0.07)

research best React state management library for 2025

→ Spawns 3 parallel agents (web + codebase + deps) → Returns comparison table + recommendation → 67% cost reduction with parallel agents

⚡ Parallel Development (1.5-3x measured speedup)

work on authentication, dashboard, and API in parallel

→ Creates development plan with task breakdown → Sets up git worktrees for parallel execution → Spawns agents to work simultaneously

💡 Feature Discovery

what can Contextune do?

→ Activates intent-recognition skill → Shows all capabilities with examples → Guides you to the right commands

📚 Most Used Commands

/ctx:research <query>

Fast technical research using 3 parallel Haiku agents.

Examples:

/ctx:research best database for user authentication
/ctx:research should I use REST or GraphQL
/ctx:research latest TypeScript testing frameworks

What you get:

  • Web research (latest trends, comparisons)
  • Codebase analysis (what already exists)
  • Dependency check (what's installed vs needed)
  • Recommendation with reasoning

Execution: Fast parallel | Cost: ~$0.07

/ctx:status

Monitor progress across all parallel worktrees.

Shows:

  • Active worktrees and their branches
  • Task completion status
  • Git commit history per worktree
  • Next steps and blockers

Use when: Working on parallel tasks and want overview

/ctx:configure

Setup guide for persistent status bar integration.

Enables:

  • Real-time detection display in status bar
  • Zero context overhead (file-based)
  • See what Contextune detected without cluttering chat

One-time setup: Adds statusline script to your config

🔧 Advanced Workflow Commands

/ctx:plan

Document development plan for parallel execution.

Creates:

  • Modular YAML plan with task breakdown
  • Dependency graph (parallel vs sequential tasks)
  • Resource allocation strategy
  • Time and cost estimates

Example:

/ctx:plan
# Then describe your features:
# "I need user auth, admin dashboard, and payment integration"

Output: .parallel/plans/active/plan.yaml

/ctx:execute

Execute development plan in parallel using git worktrees.

What happens:

  1. Reads plan from .parallel/plans/active/plan.yaml
  2. Creates git worktrees for each task
  3. Spawns parallel agents to work independently
  4. Creates PRs when tasks complete
  5. Reports progress and costs

Prerequisites:

  • Git repository with remote
  • GitHub CLI (gh) authenticated
  • Existing plan (run /ctx:plan first)

Performance: Measured speedup typically 1.5-3x on completed workflows Cost savings: 81% cheaper with Haiku agents ($0.27 vs $1.40 per workflow)

/ctx:cleanup

Clean up completed worktrees and branches.

Removes:

  • Merged worktrees
  • Completed task branches
  • Temporary files

Safe: Only cleans up merged/completed work

/ctx:stats

View detailed usage statistics and performance metrics.

Shows:

  • Detection accuracy breakdown
  • Cost savings vs manual work
  • Performance metrics (latency, throughput)
  • Most-used commands

Validates claims: See actual 81% cost reduction data

🎯 Natural Language Detection

You don't need to memorize commands! Just type what you want:

What You Type Contextune Detects Confidence
"analyze my code" /sc:analyze 85% (keyword)
"run the tests" /sc:test 85% (keyword)
"research best approach" /ctx:research 92% (keyword)
"work in parallel" /ctx:execute 88% (keyword)
"review performance" /sc:improve 85% (keyword)
"explain this code" /sc:explain 85% (keyword)

Detection tiers:

  • Keyword (0.02ms) - 60% of queries, instant
  • Model2Vec (0.2ms) - 30% of queries, fast embeddings
  • Semantic Router (50ms) - 10% of queries, LLM-based

🤖 Auto-Activated Skills

These skills activate automatically when you mention certain topics:

parallel-development-expert

Triggers: "parallel", "concurrent", "speed up", "work on multiple"

Does:

  • Analyzes if tasks are parallelizable
  • Calculates time savings
  • Suggests /ctx:plan and /ctx:execute
  • Guides worktree setup

intent-recognition

Triggers: "what can you do", "capabilities", "features", "help"

Does:

  • Shows Contextune capabilities
  • Provides natural language examples
  • Explains detection system
  • Guides to relevant commands

git-worktree-master

Triggers: "worktree stuck", "can't remove", "locked", "worktree error"

Does:

  • Diagnoses worktree issues
  • Suggests safe fixes
  • Handles lock file cleanup
  • Prevents data loss

performance-optimizer

Triggers: "slow", "performance", "optimize", "bottleneck"

Does:

  • Benchmarks workflow performance
  • Identifies bottlenecks
  • Calculates speedup potential
  • Suggests optimizations

🚀 Quick Start Workflow

1. First-Time Setup (Optional, 2 min)

/ctx:configure

→ Sets up status bar integration for persistent detection display

2. Fast Research (When You Need to Decide)

research best authentication library for Python

→ Get answer in 1-2 min with comparison table

3. Parallel Development (When You Have Multiple Tasks)

work on user auth, admin panel, and reports in parallel

→ Contextune creates plan + worktrees + executes

4. Monitor Progress

/ctx:status

→ See what's happening across all parallel tasks

5. Clean Up When Done

/ctx:cleanup

→ Remove merged worktrees and branches

💰 Cost Optimization

Contextune uses Haiku agents for 81% cost reduction:

Operation Sonnet Cost Haiku Cost Savings
Research (3 agents) $0.36 $0.07 81%
Task execution $0.27 $0.04 85%
Worktree management $0.06 $0.008 87%

Annual savings: ~$1,350 for typical usage (1,200 workflows/year)

Run /ctx:stats to see YOUR actual savings!

🔧 Configuration Files

Plan Structure

.parallel/
├── plans/
│   ├── 20251025-120000/    # Timestamped plan directories
│   │   ├── plan.yaml       # Main plan file
│   │   └── tasks/          # Individual task files
│   └── active -> 20251025-120000/  # Symlink to current plan
└── scripts/
    ├── setup_worktrees.sh  # Worktree creation script
    └── create_prs.sh       # PR creation script

Detection Data

.contextune/
└── last_detection         # JSON with latest detection
                          # Read by status bar script

📖 More Resources

🆘 Common Questions

Q: Do I need to memorize slash commands? A: No! Just type naturally. Contextune detects intent automatically.

Q: Does this slow down Claude Code? A: No. Hook adds <2ms latency for 90% of queries.

Q: Does it work offline? A: Yes! Keyword + Model2Vec tiers work offline (90% coverage).

Q: Can I customize detection patterns? A: Yes! Edit ~/.claude/plugins/contextune/data/user_patterns.json

Q: How do I see detection stats? A: Run /ctx:stats to see accuracy, performance, and cost metrics.

💡 Tip: Type "what can Contextune do?" right now to see the intent-recognition skill in action!

Reference in New Issue View Git Blame Copy Permalink