返回技能列表

express-insight

pjt222
更新于 Yesterday
6 次查看
17
2
17
在 GitHub 上查看
ai

关于

This skill translates a complex, multi-domain insight into a clear and actionable communication for a specific audience. It provides a structured method to choose the right format, convey the integrated understanding without oversimplifying, and invite constructive feedback. Use it after `integrate-gestalt` to effectively share cross-domain insights with specialists, generalists, or decision-makers.

快速安装

Claude Code

推荐
主要方式
npx skills add pjt222/agent-almanac -a claude-code
插件命令备选方式
/plugin add https://github.com/pjt222/agent-almanac
Git 克隆备选方式
git clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/express-insight

在 Claude Code 中复制并粘贴此命令以安装该技能

技能文档

Express Insight

Communicate multi-domain gestalt so it lands — preserve relationships, accessible to audience, honest where integration simplifies or risks distortion. Expression = final step synoptic cycle. Without → integrated remains private + unactionable. Challenge: language linear, insights not — provides structures for multi-dimensional communication without reducing to single dim.

Use When

  • After integrate-gestalt produces cross-domain understanding to communicate
  • Finding spans domains, single-domain framing loses relationships
  • Audience differs from perspective that generated insight
  • Integrated feels clear internally but resists straightforward expression
  • Decision depends on seeing how domains interact, not what each says independently
  • Previous attempts met w/ confusion or domain-specific pushback
  • Synoptic-mind team findings → stakeholders outside team

In

  • Required: Integrated insight (output of integrate-gestalt or equivalent)
  • Required: Audience (specialists, generalists, decision-makers, mixed)
  • Optional: Form constraints ("must fit PR description", "verbal summary")
  • Optional: Integrated domains (for attribution)
  • Optional: Prior failed attempts (what didn't land)

Do

Step 1: Assess Audience

Who + what need. Same gestalt → 3 audiences = 3 forms.

  1. Primary:
    • Specialists need domain accurate — reject oversimplification even if synthesis correct
    • Generalists need big picture — relationships matter more than details
    • Decision-makers need actionable implications + trade-offs — want what to do, cost, what if don't
    • Mixed → layered: big picture first, domain depth specialists verify
  2. Existing mental model:
    • What do they know each domain?
    • Which connections new?
    • What assumptions insight challenges?
  3. Trust requirement: how much justification before accepting cross-domain claim?
    • Specialists trust insight respecting rigor
    • Generalists trust insight making complexity navigable w/o oversimplify
    • Decision-makers trust insight surfacing trade-offs honestly not hiding

→ Clear picture who, what need, what makes credible. Assessment influences every subsequent step.

If err: audience unknown/broad → default mixed-audience: big picture + domain depth on demand. "Everyone" less effective than specific, but better than wrong guess.

Step 2: Choose Form

Form determines what audience can perceive. Not decoration.

  1. 4 primary:

    FormStructureBest for
    NarrativeStory connecting — "X in A creates Y in B, means Z"Complex/novel, audience follows reasoning
    DiagramSpatial layout — nodes = contributions, edges = connectionsStructural, topology matters more than sequence
    Comparison tableEach domain's perspective same issue parallel colsAnalytical, verify each contribution indep
    Recommendation"Do X because A, B, C converge on Y, trade-off Z"Decision-makers, need act not understand
  2. Match form to insight:

    • Causal chain across domains → narrative
    • Structural relationships → diagram
    • Convergence/divergence → comparison table
    • What to do next → recommendation
  3. Combine: recommendation backed by comparison, or narrative + diagram. But lead 1 primary — multiple formats → cognitive load obscures not clarifies

  4. Medium constraints: verbal summary can't carry table; commit msg can't carry narrative. Adjust form not force content into incompatible container

→ Primary form (+ optional secondary) w/ rationale tied to audience + insight.

If err: no form feels right → insight not yet fully integrated. Return integrate-gestalt — difficulty expressing signals incomplete synthesis, not communication.

Step 3: Express Gestalt

Communicate in chosen form, note what integrates, where simplifies, what enables.

  1. State clearly — 1-3 sentences capturing core. Gestalt itself, not evidence.
  2. Name domains — explicit which contributed. Not credit — verification. Each = invitation: "check against your expertise."
  3. Mark simplifications — every multi-domain insight simplifies:
    • Which nuances set aside?
    • Relationships treated stronger/weaker than might be?
    • What would specialist X want to add/qualify?
  4. State emergent value — what does this enable single-domain doesn't?
    • What decision possible now not before?
    • What risk visible hidden within individual domains?
    • What opportunity appears at intersection no single domain owns?
  5. Maintain multi-domain texture — resist flattening into 1 domain's language. Integrates engineering + UX → use both. Connects research + ops → preserve both framings. Texture = insight.
  6. Sequence for comprehension — non-linear insight, sequential communication. Entry point giving best foothold: start where they know, bridge outward unfamiliar. First sentence determines lean in or tune out.

→ Insight audience understands, verifies vs expertise, acts on. Simplifications visible not hidden. Value clear.

If err: feels like list of contributions not integrated whole → gestalt decomposed during communication. Step back + re-express: start from what combination reveals, not what each says separately. Synthesis = message, not parts.

Step 4: Invite Challenge

State strongest reason insight might be wrong. Integrated can feel more certain than are — synthesize many inputs → convergence creates sense of validity unearned. Not disclaimer appended for politeness; structural component making usable.

  1. Weakest link — which domain connection least well-supported? Where integration relies on analogy not evidence?
  2. Assumption at risk — what needs to be true, how confident?
  3. Counter-insight — equal access same domains, diff conclusion, strongest argument?
  4. Frame challenge as valuable — challenging strengthens. "Strongest objection I see is..." = confidence + openness simultaneously
  5. Specify what changes mind — evidence/argument revising/collapsing. Makes falsifiable not just persuasive.

→ Honest uncertainty increases not decreases trust. Insight challengeable → improvable.

If err: no weakness → warning sign. All cross-domain involve translation between frameworks, translation always loses. Loss invisible = not found not avoided. Look harder domain boundaries — hidden assumptions live there. Common hiding: shared metaphors working differently each domain, statistical correlations assumed causal across boundaries, analogies holding structurally not quantitatively.

Check

  • Audience identified, needs shaped expression
  • Form chosen by insight type + audience not habit/convenience
  • Stated as coherent whole not decomposed per-domain
  • Contributing domains named for verification
  • Simplifications stated — set aside + approximated
  • Emergent value articulated — enables vs parts
  • Multi-domain vocab preserved not flattened
  • Entry point starts where audience is, bridges to insight
  • Strongest reason wrong stated
  • Falsifiable — evidence/arguments revising named
  • Specialist reading own contribution recognizes accurate not caricature

Traps

  • Domain-by-domain reporting: Presenting each sequential ≠ insight → raw material. Insight = emerges from combination. Lead synthesis, support domain detail if needed.
  • False certainty from convergence: 3 domains same way feels strong evidence. Share underlying assumptions/data → less independent than appears. Check truly independent.
  • Flatten to audience's domain: Specialist → translate entire into their language. Accessible but destroys multi-domain. Preserve texture — unfamiliar vocab = signal not noise.
  • Skip challenge: Omit "why I might be wrong" feels stronger. Not — less trustworthy + less improvable. Epistemic honesty = feature.
  • Insight inflation: Claim synthesis reveals more than does. Cross-domain observation ≠ breakthrough. Precise scope: "applies to X in context Y" > "changes everything."
  • Premature expression: Express before fully formed → half-insights sound integrated but collapse. Stalling → upstream integrate-gestalt problem, not here.
  • Hide behind complexity: Multi-domain vocab to sound sophisticated not preserve texture. Simpler framing captures same w/o losing relationships → use simpler. Complexity necessary not performative.

  • integrate-gestalt — produces insight this expresses; express-insight = communication phase
  • argumentation — builds logical case for claim; express-insight communicates perception. Argumentation "here is why X is true"; express-insight "here is what becomes visible when you look A, B, C together"
  • teach — transfers established knowledge; express-insight conveys emergent just formed. Teach transmits; express reveals.
  • shine — channels authentic presence into communication; express-insight uses radiance carrying multi-domain w/o losing warmth/honesty
  • expand-awareness — widens perceptual field integration possible; express-insight closes cycle communicating widened
  • adaptic — meta-skill composing full synoptic cycle; express-insight = 5th + final step clear-open-perceive-integrate-express

GitHub 仓库

pjt222/agent-almanac
路径: i18n/caveman-ultra/skills/express-insight
0
agentsagentskillsai-assisted-developmentclaude-codeskillsteams

相关推荐技能

content-collections

Content Collections 是一个 TypeScript 优先的构建工具,可将本地 Markdown/MDX 文件转换为类型安全的数据集合。它专为构建博客、文档站和内容密集型 Vite+React 应用而设计,提供基于 Zod 的自动模式验证。该工具涵盖从 Vite 插件配置、MDX 编译到生产环境部署的完整工作流。

查看技能

polymarket

这个Claude Skill为开发者提供完整的Polymarket预测市场开发支持,涵盖API调用、交易执行和市场数据分析。关键特性包括实时WebSocket数据流,可监控实时交易、订单和市场动态。开发者可用它构建预测市场应用、实施交易策略并集成实时市场预测功能。

查看技能

creating-opencode-plugins

该Skill帮助开发者创建OpenCode插件,用于接入命令、文件、LSP等25+种事件。它提供了插件结构、事件API规范和JavaScript/TypeScript实现模式,适合需要拦截操作、扩展功能或自定义事件处理的场景。开发者可通过它快速构建响应式模块来增强OpenCode AI助手的能力。

查看技能

sglang

SGLang是一个专为LLM设计的高性能推理框架,特别适用于需要结构化输出的场景。它通过RadixAttention前缀缓存技术,在处理JSON、正则表达式、工具调用等具有重复前缀的复杂工作流时,能实现极速生成。如果你正在构建智能体或多轮对话系统,并追求远超vLLM的推理性能,SGLang是理想选择。

查看技能