What makes this methodology 'self-improving'?

It uses nested loops and specific completion criteria to ensure findings are refined and verified across multiple passes until they meet strict standards.

How does dialectic synthesis work within this framework?

It forces the agent to identify a thesis, actively seek a conflicting antithesis from a different source category, and then synthesize a nuanced conclusion that accounts for both.

When should I use this skill instead of a standard search?

Use it when you need factual certainty rather than just plausible answers, or when the topic is highly prone to bias and requires multiple perspectives.

Why are 'blind workers' important in this skill?

Blind workers operate without knowing the overall context of the investigation, which prevents them from simply confirming previous findings or being influenced by the orchestrator's potential biases.

Self-Improving Investigation

Name: Self-Improving Investigation
Author: bogheorghiu

bybogheorghiu

0•

学习与文档

Implements a multi-layered, self-correcting research methodology to eliminate confirmation bias and ensure factual certainty in complex investigations.

The self-improving-investigation skill provides a rigorous three-layer framework for deep-dive research within Claude Code. By utilizing an orchestration layer, blind worker agents, and dialectic synthesis, it systematically combats common AI pitfalls like confirmation bias and premature exit. It is designed for high-stakes scenarios where accuracy is more important than speed, forcing the agent to seek out antithetical viewpoints and diverse source categories before reaching a final conclusion.

主要功能

01Source diversity framework ensuring multi-perspective analysis across 6 categories

02Blind worker pattern to prevent context pollution and confirmation bias

03Nested iteration loops that refine findings until specific completion criteria are met

04Three-layer architecture comprising orchestration, blind workers, and dialectic synthesis

050 GitHub stars

06Evidence tier system for explicit labeling of verified, credible, and speculative data

使用场景

01Performing deep market or technical research requiring global source diversity

02Verifying high-stakes factual claims where bias risk and misinformation levels are high

03Systematic investigation of complex topics where single-pass research has proven insufficient

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add bogheorghiu/ex-cog self-improving-investigation

For use in Claude.ai and ChatGPT

主要功能

01Source diversity framework ensuring multi-perspective analysis across 6 categories

02Blind worker pattern to prevent context pollution and confirmation bias

03Nested iteration loops that refine findings until specific completion criteria are met

04Three-layer architecture comprising orchestration, blind workers, and dialectic synthesis

050 GitHub stars

06Evidence tier system for explicit labeling of verified, credible, and speculative data

使用场景

01Performing deep market or technical research requiring global source diversity

02Verifying high-stakes factual claims where bias risk and misinformation levels are high

03Systematic investigation of complex topics where single-pass research has proven insufficient

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add bogheorghiu/ex-cog self-improving-investigation

For use in Claude.ai and ChatGPT