Token导航 LogoToken导航TokenDH.com
研究检索权限需确认github未标认证来源可访问许可证需确认审计通过

epistemic-mapping认知映射

Agent Skill

epistemic-mapping 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

445

周安装

18

GitHub Stars

6

下载量

140
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:epistemic-mapping(认知映射)
来源仓库:https://github.com/andurilcode/skills
仓库路径:skills/epistemic-mapping
安装命令:
npx skills add https://github.com/andurilcode/skills --skill epistemic-mapping
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/andurilcode/skills --skill epistemic-mapping

简介

epistemic-mapping 用于知识状态的映射与盲点识别,区分已知、未知和潜在错误信念。

  • 适用于问题求解前的认知审计,暴露未知已知和混淆信念带来的风险。
  • 可绘制知识四象限图,评估现实真相与认知偏差之间的差距,提升决策质量。
  • 安装前应确认权限范围和维护状态,避免触发不必要的推理或外部查询,确保过程可控。
  • 适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

SKILL.md

Epistemic Mapping

Core: Before reasoning about a problem, map the quality and completeness of what you actually know. Unknown unknowns are more dangerous than known unknowns. Confusing belief for knowledge is more dangerous than both.

Distinct from Cognitive Bias Detection (audits *how* you reason) — epistemic mapping audits *what* you know and don't know.


The Knowledge Quadrants

                    YOU KNOW IT        YOU DON'T KNOW IT
                 ┌──────────────────┬──────────────────────┐
  IT'S TRUE      │  Known Knowns    │  Unknown Knowns      │
  (reality)      │  (foundation)    │  (blind spots)       │
                 ├──────────────────┼──────────────────────┤
  IT'S NOT TRUE  │  Known Unknowns  │  Unknown Unknowns    │
  (gaps)         │  (open questions)│  (surprises)         │
                 └──────────────────┴──────────────────────┘

Known Knowns — Facts you hold that are true. Foundation for reasoning. *Risk*: Mistaking belief for fact. Ask: "How do I know this? What's the evidence?"

Known Unknowns — Gaps you're aware of. Actionable — go find out.

Unknown Knowns — Things true but you don't realize you know, or don't realize they're relevant. Things the team collectively knows but individuals don't. *Risk*: Reinventing the wheel, missing relevant context. *Fix*: Wider consultation, explicit knowledge-sharing.

Unknown Unknowns — Most dangerous. Can't enumerate, but reduce by:

  • Outside perspectives
  • Examining assumptions explicitly
  • Prior failures in analogous situations
  • "What would surprise me here?"

The DIKW Stack

Knowledge quality isn't binary. Place claims on the stack:

LevelDescriptionExample
WisdomKnowing what to do with knowledge"Given X, prioritize Y"
KnowledgeSynthesized understanding"Pipeline degrades under concurrent load"
InformationInterpreted data"Latency increases 3× with 5+ parallel agents"
DataRaw observations"Latency: 340, 420, 890, 1200, 980 ms"
BeliefHeld without clear basis"Bottleneck is probably the DB"

Most analyses mix levels without labeling. Surface where you're reasoning from data vs. belief.


Pre-Analysis Questions

1. What do we know with high confidence? Direct evidence, measurement, reliable sources. "Everyone thinks" is not high confidence.

2. What do we believe but haven't validated? Assumptions held as truths. Prior experiences applied to current context. Intuitions.

3. What are our open questions? Known unknowns, ranked: which, if answered, would most change our approach?

4. Where might we have blind spots? Who's not in the room? Whose perspective are we missing? Analogous situations that surprised others?

5. What would change our minds? Pre-commit before analysis. If nothing could change your mind — that's belief, not reasoning.

6. What's the most dangerous assumption? The one whose failure most undermines the plan.


Output Format

🗺️ Knowledge Map

Known Knowns (high-confidence facts)

  • [Fact] — Source: [how we know]

Working Beliefs (assumed true, not validated)

  • [Belief] — Risk if wrong: H/M/L

Known Unknowns (open questions, ranked)

  1. [Most important gap] — Why it matters

Suspected Blind Spots

  • Areas where we might not know what we don't know
  • Outside perspectives not consulted
  • Historical analogues not examined

⚠️ Most Dangerous Assumption

  • Assumption: [Critical belief]
  • If wrong: [What breaks]
  • How to validate: [Cheapest/fastest check]
  • Cost of being wrong late: [If it surfaces after commitment]

🔄 Mind-Change Conditions

Pre-committed:

  • *"If we observe X, we revise view on Y"*
  • *"If [assumption] is false, we [response]"*

📋 Investigation Priority

Rank Known Unknowns by Importance × Feasibility:

QuestionImportanceFeasibilityPriority
Q1HighHigh1 — answer now
Q2HighLow2 — reduce incrementally
Q3LowHigh3 — opportunistic

Epistemic Hygiene

  • Label claims: Data, information, knowledge, or belief?
  • Source facts: "We know X" without source = belief in fact's clothes.
  • Pre-commit to falsifiability: Can't say what would change your mind → rationalizing, not reasoning.
  • Name the room: Who's not here? Whose knowledge are we missing?
  • Absence of evidence ≠ evidence of absence: "Haven't seen it fail" ≠ "won't fail."

Triggers

  • *"Are we treating this belief as a fact?"*
  • *"Who has relevant knowledge we haven't consulted?"*
  • *"What would a skeptic say we're missing?"*
  • *"What happened last time someone approached this without knowing X?"*
  • *"If this analysis is wrong, what's the most likely reason?"*
  • *"What question, if answered, would most change our approach now?"*

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

36.89%
按下载量换算52

Claude

28.09%
按下载量换算39

Cursor

19.05%
按下载量换算27

Gemini CLI

9.89%
按下载量换算14

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

权限需确认

当前来源未能明确判断权限范围,默认进入异常复核队列。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills