Esta página se muestra en inglés. Una traducción al español está en curso.
研究Oct 4, 2026·5 min de lectura

Research Prompt: Build a Source-Evaluation Table

A reusable research prompt that turns a list of URLs into a Markdown table of date, author, method and missing details, without browsing.

Listo para agents

Instalación lista para agent

Este activo puede instalarse después de elegir el runtime, revisar el plan y ejecutar el comando correspondiente.

Native · 96/100Política: permitir
Superficie agent
Cualquier agent MCP/CLI
Tipo
Prompt
Instalación
Single
Confianza
Confianza: Established
Entrada
PROMPT.md
Comando de instalación directa
npx -y tokrepo@latest install 3579e4d6-e42e-4f3b-9cb1-1f3800f5fdde --target codex

Ejecutar después de confirmar el plan con dry-run.

Start here

Paste your source list into any ordinary AI chat that accepts text, then paste the prompt below as your instruction. No terminal, account or API setup is needed.

What to paste: a numbered list of URLs and, when you know them, the title and site name for each. Example line: 1. https://example.com/article1 (Title: 'The Future of Remote Work', Site: TechDaily)

Where to paste it: in the same chat message as the prompt, right after the line that asks you for your sources.

How to check the output: the reply should be a Markdown table with the columns Source, Publication Date, Author/Org, Stated Method and Missing Info, followed by a short summary and a flag for opinion-style pieces. Every cell should trace back to text you supplied; unavailable fields should read Not provided, and unclear ones Unclear.

What this prompt does

It asks the assistant to build a source-evaluation table from material you provide. For each source it records publication date, author or organization, the stated research method, and what is missing. It then summarises which sources are most transparent and flags pieces whose method is not described.

The prompt relies only on the text you paste. It explicitly tells the assistant not to browse or fetch links, so treat the table as a structured reading of your own notes rather than a verification of the sources themselves.

Prerequisites and permissions

  • A plain AI chat that accepts pasted text. Nothing else is required.
  • You supply the source list; the assistant reads only what is in the conversation.
  • Because no browsing happens, the assistant cannot confirm dates, authors or methods on its own, and it must not invent them.

Limitations

  • Output quality depends entirely on the detail in your list. A URL alone usually yields several Not provided cells.
  • The assistant does not open links, so it cannot detect paywalls, changed pages or removed articles.
  • Unclear means the supplied text was ambiguous, not that the source is unreliable.
  • Do not paste private or confidential material. Anything in the chat is handled under that service's own terms, not this prompt.

Checking the result

Read the table row by row against your list: every source should appear once, and no reason should appear that you did not supply. The fictional example in the prompt (Article 1, Article 2) is illustrative only and is not a test result.

FAQ

Can it fill in a date or author I did not provide? No. The instruction is to write Not provided rather than guess, so an empty cell is expected behaviour.

Does it open my links or check if they still work? No. It works only from the pasted text and performs no external actions.

Verification note

Source reviewed; runtime not tested.

Complete reusable prompt

You are a research assistant helping evaluate sources. Your task is to create a source-evaluation table for the links I provide. I will give you a list of URLs or source descriptions (title, site name). For each source, extract or infer the following fields: publication date (if available), author or organization, stated research method (e.g., survey, experiment, case study, opinion), and a brief note on any missing information. If a field is not available from the material I provide, write 'Not provided'. Do not invent details. After filling the table, add a short summary of which sources are most transparent and which lack key details. Also, flag any source that appears to be an opinion piece without a described method. Before we begin, here is a fictional example:

INPUT EXAMPLE:

  1. https://example.com/article1 (Title: 'The Future of Remote Work', Site: TechDaily)
  2. https://example.com/article2 (Title: 'Study: Remote Work Increases Productivity', Site: Journal of Management)

OUTPUT SHAPE (example):

Source Publication Date Author/Org Stated Method Missing Info
Article 1 Not provided Not provided Not provided Date, author, method
Article 2 2023-05 Dr. Smith Survey of 500 employees None

SUMMARY: Article 2 provides clear methodology and authorship; Article 1 lacks transparency and may be opinion-based.

Now, please provide the list of sources you want evaluated. I will output the table in Markdown format. Ensure the output uses only the information I supply; if a source is inaccessible or ambiguous, note it as 'Unclear'. Review the table for consistency and completeness before sharing. Do not perform any external actions such as browsing or fetching links; rely solely on the provided text.

References and reuse

Original TokRepo prompt · CC BY 4.0. Reference documents retain their own rights.

🙏

Fuente y agradecimientos

Original TokRepo prompt, licensed CC BY 4.0. Reference context: ChatGPT release notes, reviewed 2026-10-04.

Discusión

Inicia sesión para unirte a la discusión.
Aún no hay comentarios. Sé el primero en compartir tus ideas.

Activos relacionados