gather

Collect structured findings with sources and confidence levels from code and web searches.

Updated Feb 20, 2026

Search Tags:#research #investigation #codebase #confidence #sources #websearch #findings

Ready to run

npx skills add https://github.com/supertyrelle/pelley --skill gather-supertyrelle

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: gather
Source: https://github.com/supertyrelle/pelley/tree/main/skills/gather
Command: npx skills add https://github.com/supertyrelle/pelley --skill gather-supertyrelle

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

The Gather Skill turns open-ended research questions into structured findings so you can move from scattered evidence to actionable notes quickly.

Core Features & Use Cases

Searches code first, then the web when needed for broader context.
Produces numbered findings with sources and confidence labels.
Supports research topics, codebase investigation, and refinement of earlier findings.
Use it to map an unfamiliar feature, investigate a bug or risk, or assemble evidence before distilling and ranking.

Quick Start

Use the gather skill to research authentication patterns in this codebase and return structured findings with sources and confidence levels.

Frequently Asked Questions about gather

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

FAQPage Schema

How do I collect research findings with sources and confidence levels?▼

To collect research findings with sources and confidence levels, use a structured investigation workflow that searches code first, then the web, returning numbered findings with evidence summaries and confidence ratings.

What is the best way to investigate an unfamiliar codebase feature?▼

The best way to investigate an unfamiliar codebase feature is to run a code-first discovery workflow that maps the implementation, follows up with web context, and outputs cited findings with assigned confidence levels.

How does structured research with confidence ratings work?▼

Structured research with confidence ratings works by searching local code and web sources, summarizing the evidence into concise numbered findings, and labeling each finding with a confidence score to show result reliability.

Can I expand prior research findings when investigating a bug?▼

Yes, you can expand prior research findings when investigating a bug by applying a discovery workflow that builds on existing notes, searches for new code and web evidence, and refines the output with updated sources and confidence labels.

Does codebase investigation require web search for broader context?▼

Codebase investigation requires web search for broader context only after an initial code-first search fails to provide sufficient evidence, ensuring the final structured findings include comprehensive cited sources and confidence ratings.

When should I not use an open-ended research workflow?▼

You should not use an open-ended research workflow when you need a single definitive answer rather than assembled evidence, because this process produces structured findings with confidence levels meant for distillation rather than immediate final conclusions.