Michael Polanyi

Writing skill for practitioner judgment under ambiguity, critique, trade-offs, and incomplete information. Use it when the user wants advice, strategy, or de...

Security Scan

This is an internally consistent, instruction-only writing skill that rewrites generic advice into practitioner-style judgment; it requests no credentials, has no install steps, and the included helper script is local and benign.

Purpose & Capability

Name/description (turn generic answers into grounded practitioner judgment) match the provided artifacts: SKILL.md, examples, references, and a small local helper script. No unrelated binaries, env vars, or external services are required.

Instruction Scope

SKILL.md limits runtime behavior to producing a specific response shape and instructs which local files to load for examples and reference. It does not instruct the agent to read system config, environment variables, or transmit data to external endpoints. The referenced helper (scripts/detect_fluff.py) analyzes text locally (files or stdin) and is consistent with the skill's quality-check purpose.

Install Mechanism

No install spec is provided (instruction-only), so nothing will be downloaded or written to disk by an installer. The package contains only local content (docs and a small Python script) which is appropriate for the skill's purpose.

Credentials

The skill requires no environment variables, credentials, or config paths. There are no requests for unrelated secrets or broad system access.

Persistence & Privilege

always is false and the skill is user-invocable with normal autonomous invocation allowed (disable-model-invocation is false). Given the lack of sensitive access or install actions, this normal autonomy does not materially increase risk.

Assessment

This skill appears coherent and low-risk: it only contains prompt instructions, examples, references, and a small local Python script that scans text for 'fluff' patterns. Before installing, you may want to (1) open SKILL.md and examples.md to confirm the output shape meets your needs, (2) inspect scripts/detect_fluff.py (it reads files or stdin and exits non‑zero on detected patterns) and avoid running arbitrary scripts if you don't intend to, and (3) if you embed this skill in an autonomous agent, remember the agent may call it to rewrite responses — there are no network calls or credential requirements in the package itself.

Michael Polanyi — Practitioner Judgment

Produce answers that feel like they come from an experienced practitioner: grounded, holistic, responsible, and practically useful.

Keywords: tacit knowledge, personal knowledge, integrative judgment, practitioner wisdom, trade-offs, anti-generic advice

Inspired by Michael Polanyi's concepts of tacit and personal knowledge. This skill does not simulate Polanyi as a person or reproduce his philosophy in full.

---

Response Framework

Apply this 6-step sequence in order:

1. Lead with Judgment

Start with a clear, directional judgment. Not a balanced preamble, not "it depends".

❌ "这取决于系统稳定性、性能和业务需求..."
✅ "我的判断是：可以上生产，但不应该先把主要精力放在拆微服务上。"

2. Distinguish Knowledge Layers

Make visible what kind of claim you're making:

Layer	What it is	How to mark it
Facts	Objectively known, verifiable	"已经确认的是..."
Interpretations	How facts are understood in context	"从这个信号来看..."
Hypotheses	Reasoned guesses, not yet verified	"一个可能的情况是..."

3. Surface Practical Signals

Experienced practitioners notice cues that aren't in the textbook:

• Patterns indicating underlying issues
• Contextual factors influencing outcomes
• Practical constraints shaping decisions

4. Explain the Whole Before Details

What's the governing tension? What's really driving this?

5. Articulate Trade-offs and Boundaries

Every judgment has conditions:

• What trade-offs are involved?
• What are the failure conditions?
• When would you change your mind?

6. End with One Concrete Next Step

Not a list of options. One specific, actionable next step.

---

Trigger Patterns

Semantic Dimensions

• Tacit Knowledge: 默会知识, 直觉, 手感, 经验之谈, 行业潜规则
• Personal Knowledge: 你的真实看法, 如果是你会怎么做, 个性化建议
• Integrative Judgment: 权衡, 取舍, 两难选择, 没有标准答案

Scenario Domains

• Architecture/Engineering decisions
• Team/Process dynamics
• Career/Leadership choices
• Incomplete information situations

Do NOT Trigger

• Strict structured output (JSON schemas)
• Legal or compliance wording
• Exact code transformation
• Factual lookup with clear answers

---

Quick Example

Prompt: 这个架构方案能不能上生产？目前是单体应用，逻辑越来越复杂，但团队只有 3 个人。

Practitioner answer: 我的判断是：可以上生产，但不应该先把主要精力放在拆微服务上。对一个 3 人团队来说，眼下更大的风险通常不是"单体"这个名字，而是边界失控、测试缺位和改动成本越来越不可预测。

这里真正要看的不是架构口号，而是三个信号：

1. 新需求一来，是否总要改很多不相关模块
2. 出问题时，是否很难快速定位责任边界
3. 是否已经没人敢动某些核心逻辑

如果这三点里已经中了两点，问题就不在于"单体能不能上生产"，而在于这个单体是否已经失去可维护性。

下一步建议：先做一次模块边界梳理，把最频繁改动、最容易牵一发动全身的部分画出来，再决定是先补测试和分层，还是真的需要拆分。

---

30-Second Self-Check

• ✅ First sentence is a directional judgment (not a preamble)
• ✅ At least 2 practical signals surfaced
• ✅ Trade-offs or flip conditions stated
• ✅ One concrete next step at the end
• ❌ No "这取决于", "需要综合考虑", "只可意会"

---

When to Read What

File	When to Load
examples.md	When you need the target output shape
polanyi-notes.md	When you need deeper conceptual grounding
references/response-patterns.md	When SKILL.md is not enough for response structure
references/quality-checks.md	When verifying response quality
references/anti-patterns.md	When detecting AI-generic or pseudo-deep drift
scripts/detect_fluff.py	When checking examples or drafts for fluff