consensus verification

Marketplace research is broken by SEO, affiliate marketing, and AI-generated content farms. This skill provides a systematic framework to verify if a product consensus is genuine or manufactured by analyzing search streams for hidden biases.

What it does

• Multi-stream analysis compares results from Brave Search and three Cloudflare-backed engines to detect stream-specific manipulation.
• Source classification categorizes domains into independent testers, affiliate round-ups, or suspect SEO farms based on page metadata.
• Affiliate detection scans page content for tracking links, missing disclosures, and templated press copy.
• Adversarial auditing identifies hidden text, AI-directed instructions, and marketing claims masquerading as specs.
• Criteria locking forces a weighted evaluation of candidates against measurable metrics before reviewing search results.

How it works

1. Initialize run. Call cv new to define the topic and lock measurable criteria with weighted scores.
2. Fan out search. Execute four targeted queries across all streams to find alternatives, complaints, and neutral data.
3. Inspect and digest. Analyze the top 15 results for affiliate signals and hidden text to generate a comprehensive research digest.
4. Probe and report. Score the shortlist against criteria, verify top-pick complaints, and generate a final verification report.

Frameworks & tools

The skill utilizes a custom CLI tool cv built with Python and managed via uv. It integrates with the Brave Search API and Cloudflare AI Gateway (Ceramic, Exa, Linkup).

Why this beats prompting it yourself

Standard prompting often succumbs to the "hallucination of consensus" by summarizing the most frequent search results. This skill enforces a "criteria-first" logic and uses multi-engine cross-referencing to surface what search engines or affiliate sites are intentionally omitting.

Use cases

• Product vetting before making high-value purchases like home infrastructure or electronics.
• Market research to identify if a niche is dominated by manufactured SEO consensus.
• Bias auditing of existing research to find hidden affiliate interests or bot-farm influence.
• Spec verification by separating manufacturer marketing claims from independent lab measurements.

Known limitations

Cannot detect manipulation in results no engine surfaces. Hidden-text detection is limited to inline styles and attributes, not external stylesheets. JS-heavy or bot-walled pages require fallback fetching.