# Critical Research Agent - Claude Skill

A production-grade Claude Skill that forces AI to think critically before recommending company opportunities. Built with 5-pillar triangulation and anti-hallucination validation.

The Problem
Most AI gives generic, hallucinated answers when asked "What is the opportunity to work with X company?" This damages credibility in client pitches.

The Solution
This skill transforms Claude into a critical analyst that:
- Challenges assumptions - Asks for target market, context, market conditions BEFORE answering
- Validates everything - Every number needs source + confidence tag
- Triangulates - Arithmetic, trend, source, and cross-pillar checks
- Thinks in phases - Research brief first, deck only after approval

Framework: 5 Pillars
1. Client Credentials - Corporate snapshot, momentum, SoV
2. Market & Category - Ad spend, channel mix, regulation
3. Audience Voice - Social, marketplace reviews, forum discourse. Bilingual search.
4. Variables & Geo Overlay - Specific corridors, whitespace analysis
5. Fit Assessment - Honest capability mapping

Validation System
[triangulated] = 2+ sources agree [verified] = multiple credible sources [single-source] = one source [directional] = estimated [claimed] = from subject itself

Why This Matters for AI Training
- Skill Routing & Instruction Design
- Red Teaming / Critical Thinking
- Bilingual Evaluation - Indonesian + English with local nuance
- Anti-Hallucination Architecture