Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Automatic workflow suggestion and utility agent routing based on prompt analysis. Trigger keywords: auto-detect, intent, suggest, trigger, routing, workflow detection, utility agents, prompt analysis, smart routing
.claude/skills/majiayu000-smart-intent-detection/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 198% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 61% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 24% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 12% | 0% |
The ReAcTree plugin includes smart detection that analyzes your prompts and suggests appropriate workflows or utility agents automatically.
When you type a prompt, the system:
Triggers: "find files", "where is file", "list directory", "show folder"
Fast file discovery using Glob and Grep:
**/*.rb, app/models/*.rb)Triggers: "where is defined", "find method", "find usages", "who calls"
Precise code location using LSP:
Triggers: "what changed", "show diff", "git blame", "commit history"
Git change analysis:
Triggers: "show log", "development.log", "errors in log", "slow queries"
Rails log parsing:
Triggers: implement, build, create, add, develop, user story
Suggests:
/reactree-dev - Full parallel workflow/reactree-feature - Feature-driven with user storiesStrong indicators: User story format ("As a user, I want...")
Triggers: fix, debug, error, bug, not working, troubleshoot
Suggests:
/reactree-debug - Systematic debugging workflowStrong indicators: Error types (NoMethodError), stack traces, line numbers
Triggers: refactor, cleanup, optimize, restructure, improve
Suggests:
/reactree-dev - With refactor focus and test preservationStrong indicators: "code smell", "technical debt", "decouple"
Triggers: test-first, TDD, with tests, ensure coverage
Adds TDD emphasis to feature suggestions.
Edit .claude/reactree-rails-dev.local.md:
yaml--- smart_detection_enabled: true # Enable/disable detection_mode: suggest # suggest | inject | disabled annoyance_threshold: medium # low | medium | high use_claude_analysis: true # Use Claude CLI for intelligent intent analysis ---
Shows a suggestion message. Non-intrusive - you can ignore and proceed.
Automatically activates the workflow with context injection.
Turns off smart detection entirely. Use commands manually.
Only triggers for very explicit keywords:
Skips:
Triggers for most Rails-related prompts with relevant keywords.
/reactree-*, /rails-*| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 4,318 | 12,027 | +179% | 1 | 1 | 0% | 775 | 2,311 | +198% | 0 | 0 | — |
case-02 | fail→pass | 15,568 | 6,090 | -61% | 1 | 1 | 0% | 1,709 | 2,021 | +18% | 0 | 0 | — |
case-03 | fail→pass | 8,954 | 8,554 | -4% | 1 | 1 | 0% | 1,488 | 2,401 | +61% | 0 | 0 | — |
case-04 | pass→pass | 14,096 | 8,981 | -36% | 1 | 1 | 0% | 1,329 | 1,642 | +24% | 0 | 0 | — |
case-05 | fail→pass | 6,815 | 7,582 | +11% | 1 | 1 | 0% | 1,068 | 1,320 | +24% | 0 | 0 | — |
case-06 | pass→pass | 11,898 | 2,799 | -76% | 1 | 1 | 0% | 1,015 | 1,443 | +42% | 0 | 0 | — |
case-07 | fail→pass | 12,088 | 1,951 | -84% | 1 | 1 | 0% | 1,161 | 1,302 | +12% | 0 | 0 | — |
case-08 | pass→pass | 18,631 | 4,314 | -77% | 1 | 1 | 0% | 1,818 | 1,682 | -7% | 0 | 0 | — |
case-09 | fail→pass | 25,438 | 7,283 | -71% | 1 | 1 | 0% | 3,160 | 1,332 | -58% | 0 | 0 | — |
case-10 | fail→pass | 20,951 | 15,811 | -25% | 1 | 1 | 0% | 2,310 | 2,851 | +23% | 0 | 0 | — |
case-11 | fail→pass | 32,689 | 3,663 | -89% | 1 | 1 | 0% | 4,322 | 1,555 | -64% | 0 | 0 | — |
case-12 | fail→pass | 17,060 | 3,427 | -80% | 1 | 1 | 0% | 1,643 | 1,445 | -12% | 0 | 0 | — |
case-13 | pass→pass | 12,945 | 2,437 | -81% | 1 | 1 | 0% | 1,190 | 1,360 | +14% | 0 | 0 | — |
case-14 | fail→pass | 17,259 | 17,188 | -0% | 1 | 1 | 0% | 2,659 | 2,738 | +3% | 0 | 0 | — |
case-15 | fail→pass | 16,004 | 7,289 | -54% | 1 | 1 | 0% | 1,758 | 2,137 | +22% | 0 | 0 | — |
case-16 | fail→pass | 15,788 | 5,990 | -62% | 1 | 1 | 0% | 1,977 | 1,242 | -37% | 0 | 0 | — |
case-17 | fail→pass | 19,626 | 9,026 | -54% | 1 | 1 | 0% | 1,664 | 1,615 | -3% | 0 | 0 | — |
case-18 | fail→pass | 24,530 | 9,726 | -60% | 1 | 1 | 0% | 2,264 | 1,759 | -22% | 0 | 0 | — |
case-19 | pass→pass | 14,184 | 11,555 | -19% | 1 | 1 | 0% | 2,481 | 1,982 | -20% | 0 | 0 | — |
case-20 | pass→pass | 13,348 | 23,014 | +72% | 1 | 1 | 0% | 2,430 | 3,490 | +44% | 0 | 0 | — |
case-21 | pass→pass | 14,054 | 19,351 | +38% | 1 | 1 | 0% | 1,645 | 2,274 | +38% | 0 | 0 | — |
case-22 | pass→pass | 12,526 | 17,942 | +43% | 1 | 1 | 0% | 2,145 | 3,049 | +42% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +64 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.