Run any Skill in Manus with one click

research-source-sweeper

Stars0

Forks0

UpdatedApril 25, 2026 at 11:48

Use for broad web or literature research where the user asks for many sources, source diversity, market scans, competitive research, "200 sources", evidence maps, or exhaustive discovery before synthesis. Helps plan multi-query search, deduplicate sources, preserve citations, and keep large research runs token-efficient.

Installation

Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.

Run Skill in Manus

Source

ivanglz95

ivanglz95/ai-tools-sync

View GitHub Repository View Creator Repositories

Download

Run Skill in Manus

Related occupationsSOC

Based on SOC occupation classification

News Analysts, Reporters, and JournalistsArts, Design, Entertainment, Sports, and Media Occupations·SOC 27-3023

File Explorer

2 files

SKILL.md

readonly

name	research-source-sweeper
description	Use for broad web or literature research where the user asks for many sources, source diversity, market scans, competitive research, "200 sources", evidence maps, or exhaustive discovery before synthesis. Helps plan multi-query search, deduplicate sources, preserve citations, and keep large research runs token-efficient.

Research Source Sweeper

Use this skill to turn broad research into a controlled evidence sweep.

Workflow

Define the research question, date sensitivity, geography, languages, and source types.
Split into focused queries. Prefer 10-20 precise queries over one vague query.
Use the best available source tools:
- Built-in web search for current facts and general discovery.
- Exa, Tavily, Brave Search, or similar MCP tools when installed and API keys are configured.
- Context7 for current software documentation, not general web search.
- GitHub/search APIs for repositories and issues.
Collect only compact metadata first: title, URL, date, publisher, snippet, relevance, and query that found it.
Deduplicate by canonical URL, normalized title, and near-identical snippet.
Rank sources by relevance, authority, recency, and independence.
Fetch full pages only for the shortlist needed to answer.
Return a synthesis with citations and a brief "coverage notes" section naming what was searched and what was not.

High-Volume Targets

For requests like "use 200 sources", do not paste 200 source excerpts into context. Build a source index first, then summarize in layers:

Discovery target: up to 200 candidate URLs.
Evidence shortlist: 20-40 high-quality sources.
Deep-read set: 5-15 sources unless the user asks for a formal bibliography.

When a tool has low per-query result limits, page or batch queries. If the installed search tools cannot reach the requested volume, say exactly what limit was hit and recommend a provider such as Exa, Tavily, or Brave Search MCP.

Output Pattern

For normal answers:

Direct answer.
Key evidence with citations.
Source coverage summary.
Gaps or uncertainty.

For research handoffs, create a compact table with:

Keep notes short. Avoid large quotations unless the user asks.

More from this repository

same repository

context-clarifier

ivanglz95/ai-tools-sync

ALWAYS use at the start of every user request, new chat, project task, research task, coding task, troubleshooting session, document/spreadsheet/presentation task, automation, or open-ended question to preserve context, ask only necessary clarifying questions, identify missing inputs, select relevant skills/plugins/model tier, and improve answer quality without bloating the context window.

2026-04-260

project-tool-advisor

ivanglz95/ai-tools-sync

ALWAYS use at the beginning of every new chat, project, coding task, repository setup, automation request, research task, codebase task, troubleshooting task, document/spreadsheet/presentation task, or whenever the user asks what skill/plugin/tool/model should be used. Performs a mandatory preflight that recommends relevant installed or missing skills/plugins/MCPs and a model tier before heavy work starts, while checking the sync workflow first.

2026-04-260

token-budget-manager

ivanglz95/ai-tools-sync

Use when a task may consume a large context window or many tokens: long chats, large codebases, repeated file reads, big research tasks, multi-step debugging, migrations, refactors, or when the user asks to save tokens, work cheaply, compact context, summarize progress, or avoid unnecessary reading.

2026-04-260

argi-consultor-impositivo-contable-laboral

ivanglz95/ai-tools-sync

Consultor impositivo-contable-laboral especialista en Argentina. Activar SIEMPRE ante consultas sobre AFIP/ARCA, CUIT, IVA, Ganancias, Bienes Personales, Monotributo, Ingresos Brutos, Convenio Multilateral, sueldos, cargas sociales, balances, ajuste por inflación, estados contables, sociedades comerciales, liquidaciones impositivas, planes de pago, moratorias, Libro IVA Digital, formulario 931, SIPA, RT FACPCE, inscripción en AFIP/ARCA, recategorización, exclusión de monotributo, pase a responsable inscripto, vigencia de normas, orden de prelación normativa, Ley de Inocencia Fiscal, RT 54 NUA, RG 5803/2025, o cualquier tema fiscal/contable/laboral argentino.

2026-04-250

file-context-scout

ivanglz95/ai-tools-sync

Use for file-heavy or codebase-heavy tasks where the user wants to save tokens or avoid reading too much: finding relevant files, understanding a repo, editing large projects, auditing configs, comparing generated artifacts, or working with large logs, docs, PDFs, spreadsheets, or source trees.

2026-04-250

name	research-source-sweeper
description	Use for broad web or literature research where the user asks for many sources, source diversity, market scans, competitive research, "200 sources", evidence maps, or exhaustive discovery before synthesis. Helps plan multi-query search, deduplicate sources, preserve citations, and keep large research runs token-efficient.

Research Source Sweeper

Use this skill to turn broad research into a controlled evidence sweep.

Workflow

Define the research question, date sensitivity, geography, languages, and source types.
Split into focused queries. Prefer 10-20 precise queries over one vague query.
Use the best available source tools:
- Built-in web search for current facts and general discovery.
- Exa, Tavily, Brave Search, or similar MCP tools when installed and API keys are configured.
- Context7 for current software documentation, not general web search.
- GitHub/search APIs for repositories and issues.
Collect only compact metadata first: title, URL, date, publisher, snippet, relevance, and query that found it.
Deduplicate by canonical URL, normalized title, and near-identical snippet.
Rank sources by relevance, authority, recency, and independence.
Fetch full pages only for the shortlist needed to answer.
Return a synthesis with citations and a brief "coverage notes" section naming what was searched and what was not.

High-Volume Targets

For requests like "use 200 sources", do not paste 200 source excerpts into context. Build a source index first, then summarize in layers:

Discovery target: up to 200 candidate URLs.
Evidence shortlist: 20-40 high-quality sources.
Deep-read set: 5-15 sources unless the user asks for a formal bibliography.

Output Pattern

For normal answers:

Direct answer.
Key evidence with citations.
Source coverage summary.
Gaps or uncertainty.

For research handoffs, create a compact table with:

Keep notes short. Avoid large quotations unless the user asks.