菜单
Metacognitive know/don't-know assessment before deep research or complex implementation. Surfaces knowledge gaps to prevent wasted exploration cycles.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Scan for stale memory topics, near-dead instincts, and low-salience K-LEAN entries. Proposes cleanup actions interactively.
Quality gate for instinct persistence. Validates new instincts through a 4-stage pipeline before allowing them to be written to disk. Invoke before writing any new instinct file.
Use when the user needs thorough research on a complex topic, says 'deep dive', 'research this thoroughly', or 'sensor sweep'. Parallel agent orchestrator for comprehensive multi-angle investigation.
Autonomous experiment loop: iteratively improves a single file against a measurable metric. Uses git-as-state-management in a disposable worktree. Invoke when user wants to optimize, tune, or systematically improve a file through repeated iterations.
Use when you need multi-faction review combining security, strategy, cost, and quality perspectives. Trigger: fleet review, multi-faction assessment, comprehensive quality gate, or when escalation requires L5+.
Two-loop research orchestrator: chains deep-research (outer, hypothesis generation) with experiment-loop (inner, empirical testing). Use when research questions have a testable dimension with a measurable metric.
| name | confidence |
| description | Metacognitive know/don't-know assessment before deep research or complex implementation. Surfaces knowledge gaps to prevent wasted exploration cycles. |
Structured assessment of what you know vs. what you're assuming vs. what you don't know about the current task. Run before /deep-research, complex multi-file implementation, or any task where wrong assumptions could waste a phase.
State the task in one sentence. Then list the 3-5 key technical questions the task depends on.
For each question, classify into exactly one category:
| Question | State | Evidence | Gap-Closing Action |
|----------|-------|----------|-------------------|
| [question] | KNOWN | [file:line, doc URL, or K-LEAN entry] | - |
| [question] | ASSUMED | [reasoning basis, no hard evidence] | [verify how] |
| [question] | UNKNOWN | - | [research action needed] |
Rules:
Confidence Score: [KNOWN count] / [total questions] = [percentage]%
Risk Level: HIGH (< 50%) | MEDIUM (50-75%) | LOW (> 75%)
If Risk Level is HIGH:
If Risk Level is MEDIUM:
If Risk Level is LOW:
Present the assessment table + risk level, then ask:
Confidence: [X/Y] ([percentage]%)
Risk: [HIGH/MEDIUM/LOW]
Unknowns: [count] requiring action
Assumptions: [count] requiring verification
Proceed? [Y / research unknowns first / verify assumptions first]
From Compound Knowledge Plugin fleet review (2026-03-24): /kw:confidence is a genuine gap in our setup. We have no explicit "what do I know vs. what don't I know" assessment. Instincts encode past knowledge; fleet reviews evaluate output; but nothing surfaces uncertainty about the CURRENT task in real time. This prevents research rabbit holes (exploring what you already know) and implementation failures (building on unverified assumptions). Ferengi ROI: 5-minute assessment prevents 30-60 minutes of wasted exploration.