Structured Thinking¶
Source: hooks/ways/meta/think/think.md
Frontmatter
| Field | Value |
|---|---|
description |
structured reasoning, thinking frameworks, cognitive scaffolding for complex decisions |
vocabulary |
explore options approaches trade-off balance alternatives stuck principle abstract reasoning framework systematic |
pattern |
explore.{0,30}(option|approach|alternativ)|weigh.{0,30}(option|trade|alternativ)|trade.?off|several (option|approach|alternativ|way)|competing (objective|priorit|concern|goal)|first principle|step.?back|i'?m stuck|which (approach|option) |
refire |
0.15 |
scope |
agent, subagent |
When you encounter complexity, don't reach for a framework first. Evaluate whether you need one.
The Metacognitive Check¶
Before solving, pause and assess: is your understanding trending toward clarity or away from it?
Do not attempt to solve in this first cycle. Just evaluate the direction:
- Trending clear — You can see the shape of the answer. Proceed normally. No scaffolding needed.
- Trending unclear — The problem has competing concerns, hidden dependencies, or you're uncertain which direction to go. Escalate.
Escalation Gradient¶
| Level | What happens | When |
|---|---|---|
| Internal reasoning | Think harder silently — extend your reasoning, consider more angles | Unclear but likely resolvable with more thought |
| External strategy | Use a structured strategy (below) — surfaces your reasoning step-by-step | Internal reasoning isn't converging; the human should see the work |
| Collaborative | Discuss with the human — they have context you lack | Strategy hits unknowns that tools can't resolve |
Most problems resolve at level 1. The strategies exist for when they don't.
External Strategies¶
When you decide to escalate, act immediately. Invoke the skill — don't announce your intention, don't ask permission, don't hedge with "I might want to use..." The decision to escalate IS the decision to act. The human cannot follow your reasoning speed; by the time they'd read a proposal to use a strategy, you should already be working through it.
| Problem Shape | Strategy | Invoke |
|---|---|---|
| Multiple viable approaches | Tree of Thoughts | /think tree |
| Three competing objectives | Trilemma | /think trilemma |
| High-stakes, need confidence | Self-Consistency | /think consistency |
| Stuck, need first principles | Step-Back | /think stepback |
| Investigation or debugging | ReAct | /think react |
Each strategy is a step-by-step scaffold that surfaces your reasoning visibly.
Engaging the Human¶
During any strategy, if you encounter unknowns that your tools can't resolve, use AskUserQuestion to engage the human. Your job is to frame the unknown precisely — not "what should I do?" but:
- "I've narrowed this to X or Y. Here's what I can't determine — which is true in your context?"
- "This depends on [specific constraint] that I don't have visibility into."
- "I've identified three approaches with these trade-offs. Which trade-off is acceptable?"
Your reasoning speed and breadth does the framing work. The human's context and ground truth fills the gap. That exchange — the agent framing, the human resolving — is the actual collaboration.
See Also¶
- trust/autonomy(meta) — thinking strategies support autonomous reasoning