Counter-Evidence Hunter is a specialized AI agent skill designed to act as a digital bloodhound, actively seeking out conflicting evidence and alternative explanations to stress-test claims.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install counter-evidence-hunter
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install counter-evidence-hunter using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Counter-Evidence Hunter?
Counter-Evidence Hunter is a high-precision diagnostic skill for AI agents that mitigates confirmation bias and reduces hallucinations. In the ecosystem of Openclaw Skills, this tool functions as a professional skeptic. It does not rewrite your primary narrative; instead, it identifies vulnerabilities by hunting for direct contradictions (counter_evidence), conditions that would flip a conclusion (flip_conditions), and data supporting alternative pathways (alternative_supports).
By systematically breaking down a mainline claim into verifiable sub-propositions, this skill ensures that high-stakes decisions are informed by a 360-degree view of available data. It is an essential component for any workflow requiring rigorous risk analysis, investigative research, or objective policy evaluation.
Counter-Evidence Hunter Use Cases
- Validating high-stakes strategic claims before final decision-making.
- Reducing single-line narrative bias in investigative research and data analysis.
- Stress-testing overconfident AI outputs to identify potential hallucinations.
- Identifying dynamic 'assumption cracks' when unexpected findings surface during a search.
- Evaluating the robustness of technical or policy conclusions through rigorous cross-examination.
How Counter-Evidence Hunter Works
- Deconstruction: The skill analyzes the
mainline_claimand breaks it into multiple independent sub-propositions for verification. - Strategic Query Generation: It designs at least three distinct counter_queries per sub-proposition to target different types of evidence.
- Dynamic Adaptation: If
unexpected_findingsare provided, it triggers specialized queries for 'assumption cracks' or 'overconfident signals'. - Evidence Classification: It executes searches and classifies results into hard evidence, soft evidence, or noise.
- Flip Condition Extraction: From the gathered evidence, it extracts structured conditions that would theoretically invalidate the primary claim.
- Confidence Assessment: It calculates a weighted confidence score based on source quality, coverage, freshness, and evidence sufficiency.
Counter-Evidence Hunter Setup
To integrate this skill into your AI agent environment, ensure the following configuration is applied within your Openclaw Skills directory:
# Navigate to your skills library
cd openclaw-skills/research-tools
# Initialize the Counter-Evidence Hunter module
openclaw install counter-evidence-hunter
# Ensure reference documents are in place
touch references/counter-patterns.md references/flip-condition-examples.md
Counter-Evidence Hunter Data Schema & Taxonomy
The skill utilizes a structured JSON schema to handle inputs and outputs. Key data structures include:
| Object | Description |
|---|---|
counter_evidence |
Contains the content, source, and strength rating (hard, soft, noise). |
flip_conditions |
Defines conditions with probability and impact_if_triggered. |
confidence_assessment |
A multi-dimensional score (0-100) based on five weighted criteria. |
alternative_supports |
Maps paths that contradict, qualify, or extend the mainline claim. |
Counter-Evidence Hunter Advanced Features
- Dynamic Discovery Hooks: Automatically generates new search directions based on
assumption_crackornew_dimensiontriggers. - Weighted Scoring V2: Uses a complex formula: $source_quality * 0.25 + coverage * 0.20 + freshness * 0.20 + counter * 0.20 + consistency * 0.15$.
- Multi-Level Extraction: A three-tier workflow that moves from raw queries to distilled flip conditions.
- Strict Red-Line Enforcement: Built-in safeguards prevent the agent from returning biased results or making final adjudications, preserving objectivity.
SKILL.md
Loading
Related Openclaw Skills
SellerSprite Product Research
A powerful command-line interface for SellerSprite that enables automated Amazon market research, keyword analysis, and competitor tracking.
Obsidian Vault Context
A skill that transforms an Obsidian vault into a shared, persistent workspace where AI agents and users collaborate via markdown files rather than just chat.
Kandy
Kandy is a powerful real-time communications platform integration that enables AI agents to manage voice, video, and messaging workflows seamlessly.
Retail Sales Performance Analysis
A professional business intelligence skill for retail environments that provides automated period-over-period sales analysis and performance diagnostics.
Skill Reviewer
A professional quality assurance tool designed to audit AgentSkill files for logical integrity, token efficiency, and structural clarity.
Telegram PDF Segregator
An automated workflow to download, filter, and organize PDF documents from Telegram channels into structured local folders.
Featured*
