gather-evidence

Collect and normalize product evidence from connected sources into weighted items.

14|Updated Jun 29, 2026

Search Tags:#evidence #identity resolution #support tickets #signal analysis #customer feedback #product research

Ready to run

npx skills add https://github.com/sparkline-ventures/product-eval --skill gather-evidence-sparkline-ventures

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: gather-evidence
Source: https://github.com/sparkline-ventures/product-eval/tree/main/skills/gather-evidence
Command: npx skills add https://github.com/sparkline-ventures/product-eval --skill gather-evidence-sparkline-ventures

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill helps you gather real product evidence from connected sources, uploads, and the web so you can validate problems with facts instead of anecdotes.

Core Features & Use Cases

Source retrieval: Pull support tickets, CRM notes, issue trackers, analytics, reviews, and other product signals from available connectors.
Evidence normalization: Turn raw items into structured evidence with claims, personas, funnel stages, timestamps, links, and strength ratings.
Identity resolution and weighting: Deduplicate the same person or account across sources, assign confidence, and weight evidence by strength and recency.
Use case: Before deciding what to build next, use this Skill to collect recurring customer pain from multiple systems, identify the strongest patterns, and hand the results to synthesis or decision-making workflows.

Quick Start

Use the gather-evidence skill to collect and normalize product evidence for the current scope.

Frequently Asked Questions about gather-evidence

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I collect product evidence from support tickets and CRM records to validate customer pain?▼

To collect product evidence, normalize raw items from connected sources like support tickets and CRM records into structured claims with personas, timestamps, and strength ratings. Apply identity resolution to deduplicate accounts and weight evidence by recency for downstream ranking.

What is evidence normalization for product research and how does it work?▼

Evidence normalization transforms raw signals from analytics, reviews, and issue trackers into structured items containing claims, personas, funnel stages, and links. It applies signal filtering and freshness checks to ensure only verified, weighted evidence items pass to decision workflows.

How do I deduplicate customer feedback across multiple systems before prioritizing features?▼

Deduplicate customer feedback across systems by applying identity resolution to match the same person or account across connected sources. Assign confidence scores and weight each evidence item by strength and recency to identify the strongest recurring pain patterns.

Does this approach work with unstructured file uploads and web sources for signal analysis?▼

Yes, this approach works with unstructured file uploads and web sources by applying source probing and signal filtering. It extracts recurring customer pain from available connectors and normalizes the raw web data into structured, scored evidence items.

What's the best way to score and weight evidence items from issue trackers for decision workflows?▼

The best way to score evidence items is to apply freshness checks and assign confidence ratings during normalization. Weight each item by signal strength and recency, then pass the structured, weighted evidence directly to downstream synthesis and ranking workflows.

When should I not use automated evidence gathering for product research?▼

You should not use automated evidence gathering when source probing cannot reliably access your target systems, or when raw items lack sufficient metadata for identity resolution and signal filtering. It requires connected sources or structured uploads to produce weighted, normalized evidence.