Bỏ qua tới nội dung
TakoAPI
Quay lại kỹ năng

curiosity-engine

bởi luofulily1-cmyk

Coding Agents & IDEs
325Lượt tải do clawskills.sh công bố
1Lượt xem
0 lượt thíchĐăng nhập để thích kỹ năng này
—Chưa có đánh giá
Đánh giá của bạn

Tóm tắt

Curiosity-driven reasoning enhancement for OpenClaw agents.

Mô tả

Kỹ năng này làm gì

Curiosity Engine reshapes how an agent reasons at inference time using a structured OODA-C loop: Observe, Orient, Doubt, Act, and Curiose. It injects self-questioning, assumption challenging, and information gap detection before the agent responds. No training required — it operates as a reasoning protocol during live sessions.

Unlike prompt-level instructions to 'think step by step', this skill provides a concrete, repeatable loop with defined protocols for doubt and gap detection, making deeper reasoning reproducible across sessions.

Khi nào nên dùng

  • Researching an unfamiliar technical topic before answering
  • Investigating why a system behaves unexpectedly
  • Digging deeper into a surprising data finding
  • Fact-checking assumptions in a complex analysis
  • Exploring open-ended questions where quick answers fall short

Quy trình ví dụ

ĐẦU VÀO

User asks: 'Why did our API latency spike last Tuesday?'

AGENT
  1. 1

    Observe available logs, metrics, and context provided

  2. 2

    Form initial hypothesis about likely causes and rate confidence

  3. 3

    Run Protocol A: generate 3 unexplored questions about the spike

  4. 4

    Run Protocol C: map known facts vs assumed vs missing data

  5. 5

    Use web_search or exec to fill highest-priority knowledge gaps

  6. 6

    Reflect on findings, update confidence, log open threads

ĐẦU RA

Structured response with findings, updated confidence rating, any surprises noted, and open threads for follow-up

Cài đặt

clawhub install luofulily1-cmyk/curiosity-engine

Chạy lệnh CLI, hoặc dán prompt vào Claude Code, Cursor, Codex hay coding agent bạn đang dùng.

Liên kết