Adaptive Reasoning
by @marjoriebroad
Automatically assess task complexity and adjust reasoning level. Triggers on every user message to evaluate whether extended thinking (reasoning mode) would...
clawhub install mar-adaptive-reasoningπ About This Skill
name: adaptive-reasoning description: Automatically assess task complexity and adjust reasoning level. Triggers on every user message to evaluate whether extended thinking (reasoning mode) would improve response quality. Use this as a pre-processing step before answering complex questions.
Adaptive Reasoning
Self-assess complexity before responding. Adjust reasoning level dynamically.
Quick Assessment (run mentally on every request)
Score the request 0-10 on these dimensions:
| Signal | Weight | Examples | |--------|--------|----------| | Multi-step logic | +3 | Planning, proofs, debugging chains | | Ambiguity | +2 | Nuanced questions, trade-offs, "it depends" | | Code architecture | +2 | System design, refactoring, security review | | Math/formal reasoning | +2 | Calculations, algorithms, logic puzzles | | Novel problem | +1 | No clear pattern, requires creativity | | High stakes | +1 | Production changes, irreversible actions |
Subtract:
Decision Thresholds
| Score | Action |
|-------|--------|
| β€2 | Stay fast. No reasoning needed. |
| 3-5 | Standard response. Light internal deliberation. |
| 6-7 | Consider /reasoning on or pause to think explicitly. |
| β₯8 | Activate extended thinking. Complex problem warrants it. |
Activation (Automatic)
Do not ask. Just activate.
| Score | Action | |-------|--------| | β€5 | Respond normally. No change. | | 6-7 | Enable reasoning silently. Add π§ at end of response. | | β₯8 | Enable reasoning. Add π§ π₯ at end of response. |
Visual Indicator
Always append the reasoning icon at the very end of your response:
π§ (thinking mode active)π§ π₯ (deep thinking mode)How to Activate
Use session_status tool or /reasoning on command internally before responding:
/reasoning on
Or via tool:
{"action": "session_status", "reasoning": "on"}
After completing a complex task, optionally disable to save tokens on follow-ups:
/reasoning off
Examples
Low complexity (score: 1) > "What time is it in Tokyo?" β Simple lookup. Answer immediately. No icon.
Medium complexity (score: 4) > "Refactor this function to be more readable" β Standard response with brief explanation. No icon.
High complexity (score: 7) > "Design a caching strategy for this API with these constraints..." β Enable reasoning. Thoughtful response ends with: π§
Very high complexity (score: 9) > "Debug why this distributed system has race conditions under load" β Enable extended thinking. Deep analysis ends with: π§ π₯
Integration
This skill runs as mental preprocessing. No external tools needed.
For explicit control:
/reasoning on β Enable extended thinking/reasoning off β Disable (faster responses)/status β Check current reasoning stateWhen NOT to Escalate
Auto-Downgrade
After completing a complex task (score β₯6), if the next message is simple (score β€3):
π‘ Examples
Low complexity (score: 1) > "What time is it in Tokyo?" β Simple lookup. Answer immediately. No icon.
Medium complexity (score: 4) > "Refactor this function to be more readable" β Standard response with brief explanation. No icon.
High complexity (score: 7) > "Design a caching strategy for this API with these constraints..." β Enable reasoning. Thoughtful response ends with: π§
Very high complexity (score: 9) > "Debug why this distributed system has race conditions under load" β Enable extended thinking. Deep analysis ends with: π§ π₯