Example-first command reference and quick start guide
Shows Contextune's command reference and quick start guide with copy-paste examples for parallel development, research, and natural language detection. Use when you need to discover capabilities or learn commands.
/plugin marketplace add Shakes-tzd/contextune/plugin install contextune@ContextuneJust type naturally—Contextune detects your intent automatically!
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
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
what can Contextune do?
→ Activates intent-recognition skill → Shows all capabilities with examples → Guides you to the right 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:
Execution: Fast parallel | Cost: ~$0.07
/ctx:statusMonitor progress across all parallel worktrees.
Shows:
Use when: Working on parallel tasks and want overview
/ctx:configureSetup guide for persistent status bar integration.
Enables:
One-time setup: Adds statusline script to your config
/ctx:planDocument development plan for parallel execution.
Creates:
Example:
/ctx:plan
# Then describe your features:
# "I need user auth, admin dashboard, and payment integration"
Output: .parallel/plans/active/plan.yaml
/ctx:executeExecute development plan in parallel using git worktrees.
What happens:
.parallel/plans/active/plan.yamlPrerequisites:
gh) authenticated/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:cleanupClean up completed worktrees and branches.
Removes:
Safe: Only cleans up merged/completed work
/ctx:statsView detailed usage statistics and performance metrics.
Shows:
Validates claims: See actual 81% cost reduction data
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:
These skills activate automatically when you mention certain topics:
parallel-development-expertTriggers: "parallel", "concurrent", "speed up", "work on multiple"
Does:
/ctx:plan and /ctx:executeintent-recognitionTriggers: "what can you do", "capabilities", "features", "help"
Does:
git-worktree-masterTriggers: "worktree stuck", "can't remove", "locked", "worktree error"
Does:
performance-optimizerTriggers: "slow", "performance", "optimize", "bottleneck"
Does:
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
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!
.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
.contextune/
└── last_detection # JSON with latest detection
# Read by status bar script
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!