Preface

In DeepSeek Harness or other SKILL.md-compatible agent environments, when addressing complex social, economic, or systemic problems, models often fall into single narratives or unverified conclusions. deep-structural-analysis-skill provides a structured analysis framework designed to improve the depth and credibility of outputs through multidisciplinary cross-validation and counter-evidence mechanisms.

What It Is

This is a cross-disciplinary deep structural analysis skill. It is maintained by tetckx, and the current version is v1.9.5. It standardizes the analysis workflow for complex problems into: Decomposition → Research → Multi-lens Analysis → Structural Tool Application → Synthesis and Validation (including cross-validation, confidence labeling, and blind-spot labeling).

This skill does not directly provide information lookup; instead, it provides an analytical workflow that ensures all conclusions are grounded in facts, with explicit confidence levels and cognitive blind spots labeled.

Core Features

The skill includes the following core mechanisms and components:

  • Analysis Protocol: Includes the “Attack Cycle Protocol” and the “Inference Four Checks”. The Attack Cycle Protocol requires exposing the default position (target) before analysis and attacking opposing positions before delivery, countering the tendency toward single narratives. The Inference Four Checks covers elimination of competing hypotheses, second-order effect inference, physical anchoring checks, and consideration of Global South variables.
  • Lens System: Provides 16 lenses divided into four categories:
    • Foundation: epistemology, systems theory, history, temporality
    • Human: psychology, sociology, anthropology, emotion
    • Structure: economics, political science, institutional analysis, technology research, geography
    • Material (domain-triggered): ecology, infrastructure, life sciences
  • Tool Pool: Built-in 10 core tools, including three-way perspective, 80-20 rule, adaptive cycles, path dependence, asymmetry detection, incentive mapping, capital matrix, reflexivity, multi-layer signal decoding (MLSD), and strategic interaction matrix.
  • Quality Standards: Enforces confidence assessment (high confidence requires ≥2 independent sources with differing positions), prohibits false balance, and requires the historical lens to be grounded in facts.

Installation and Enablement

The skill is provided in standard SKILL.md format and has no external configuration files. To install it, copy the files to the corresponding directory.

For DeepSeek Harness, perform the following steps:

  1. Create (or navigate to) the .agents/skills/ directory in your project root.
  2. Create the subdirectory deep-structural-analysis.
  3. Place the skill files (SKILL.md and related documents) in that directory.

After installation, users can start using it by specifying the skill in their prompt or triggering it directly in the conversation.

Typical Usage

This skill is suitable for scenarios that require deep exploration of structural causes.

  1. Deep Analysis: The user enters, “Do a deep analysis of the current employment situation.”
  2. Multi-angle Exploration: The user asks, “Analyze this phenomenon from multiple angles.”
  3. Framework Application: The user asks, “Use the three-way perspective to examine this issue.”
  4. Structural Attribution: The user asks, “Why is labor law enforcement so difficult?”

The skill outputs an analysis report according to the preset workflow, including modules such as “Exposed Target,” “Core Findings,” “Multi-lens Consensus,” “Confidence,” “Revision Trace,” “Layered Impact,” and “Blind Spots.”

Example output format:

深度声明：Standard · 复杂度域 Complex

【暴露靶子】默认立场："地铁持续亏损 = 应该停止扩建"（待攻击）
【核心发现】"亏损"是会计口径产物（高折旧 + 公益票价法定），扩建的
  隐性收益（土地增值/城市密度/通勤成本）不在报表内
  → 亏损叙事 vs 全成本收益错位，而非"该不该建"的问题
【多镜共识】经济学（口径 vs 全成本核算）/ 制度（公益票价法定）/
  物理（客流密度约束）
【置信度】中——"会计口径"是事实（高），"隐性收益主导决策"是解释（中）
【修订痕迹】反证"地方债务压力应叫停扩建"被评估：债务约束真实，
  但被土地出让收益部分对冲 → "叫停"结论降级为"节奏放缓"
【分层影响】系统（城市蔓延模式）/ 制度（补贴机制）/ 个体（通勤者成本）
【盲点】地方财政明细不可见；人口流出城市的个案未覆盖

Suitable Scenarios and Notes

This skill is suitable for complex social, economic, philosophical, or systemic issues that require exploration of structural causes.

Not suitable for:
* Factual explanation questions (e.g., “Why is the sky blue?”).
* Program debugging or simple information lookup.
* Content summarization.

Notes:
* Output language follows the user’s query language (Chinese query → fully Chinese output; English query → fully English output).
* Depth defaults to Standard; users can specify it in the prompt.
* Ensure it runs under the current DSH process permissions; before installation, it is recommended to check the source code and license (MIT).

Summary

deep-structural-analysis-skill provides agents with a rigorous structured analysis tool by introducing mandatory lens discipline, the Attack Cycle Protocol, and confidence calibration. It does not directly generate conclusions; rather, it presents the analytical process, helping users examine problems from multidisciplinary perspectives and reduce cognitive blind spots.