agentleFS
teodstrteodstr/garage/skills/desk-research/SKILL.md

desk-research

Skill0 starsChanged 2 days agoMITView on GitHub

Run rigorous desk research on a market, an industry or a customer group: turn the topic into precise questions, search official statistics, filings, industry and academic sources, grade every source, cross-check the key numbers, and deliver a sourced brief with dates and confidence levels. Use for market research, industry overviews, finding data for a business plan, or checking a figure. Triggers: "market research", "desk research", "find data", "industry report", "statistics", "étude de marché", "recherche documentaire", "trouver des chiffres", "données de marché", "estudio de mercado", "investigación documental", "datos del sector".

4 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Before interviews, to understand a market and its vocabulary.

    line 21
  2. Very specific:

    Not to estimate a market size from scratch: use market-sizing from boardroom if it is installed, with the figures found here.

    line 24
  3. Distinctive:

    Every number has a source, a date and a confidence level.

    line 63

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Desk research

The hard part of desk research is knowing how much to trust a number. This skill plans the search, goes to the sources closest to the data, grades each one, and shows the founder what is solid and what is still a guess.

When to use

  • Before interviews, to understand a market and its vocabulary.
  • To fill a business plan, a grant application or a pitch deck with sourced figures.
  • To check a figure someone quoted.
  • Not to estimate a market size from scratch: use market-sizing from boardroom if it is installed, with the figures found here.

Inputs

  1. The questions to answer, and the decision behind them.
  2. Geography, period and segment.
  3. Sources the founder already has.

Method

  1. Turn the topic into precise questions, each with its unit, geography and year: "number of independent bakeries in France in 2024", not "the bakery market".
  2. Search in this order, closest to the data first:
    • Official statistics: national institutes (INSEE in France, INE in Spain, their equivalents elsewhere), Eurostat, OECD, World Bank, central banks.
    • Filings and registries: annual reports, company accounts (in France, for example, Infogreffe or Pappers), stock market filings.
    • Trade associations, regulators and public agencies.
    • Academic research (for example Google Scholar) for behaviors and definitions.
    • Market research firms and the press: useful, but check how the figure was produced, and whether the article only repeats the summary of a paid report.
    • Signals: search trends, app store rankings, job postings, customer reviews, forums.
  3. Grade every source: who produced it, with which method and sample, when, and whether the method is public. Give a confidence level: high (official or audited), medium (reputable, method partly public), low (unknown method, marketing material, unsourced summary, including summaries written by AI).
  4. Cross-check: for each key figure, find a second independent source or a bottom-up sanity check. Explain the gaps between sources (different definitions, years or scopes).
  5. Cite precisely: title, publisher, publication date, link and date accessed. Without web access, say which figures must be checked and where, instead of quoting from memory.
  6. Write the brief: answers first, then the evidence, then what is still unknown and how to find it (often with customer-interviews or survey-design).

Output

### Research brief: <topic> (<geography>, <period>)
| Question | Answer | Source (publisher, date, link) | Accessed | Confidence |
|---|---|---|---|---|

### Cross-checks and gaps
…

### Still unknown, and how to find out
…

Quality bar

  • Every number has a source, a date and a confidence level; nothing is quoted from memory.
  • Primary and official sources come first; a press article that quotes a study is traced back to the study.
  • Definitions and years match when figures are compared.
  • Unknowns are listed, not filled with guesses.
  • Written in the user's language, with the terms below.

Terminology

EnglishFrançaisEspañol
Desk researchRecherche documentaireInvestigación documental
Market researchÉtude de marchéEstudio de mercado
Primary, secondary sourceSource primaire, secondaireFuente primaria, secundaria
Official statisticsStatistique publiqueEstadística oficial
Cross-checkRecoupementContraste de fuentes
Confidence levelNiveau de fiabilitéNivel de fiabilidad

Sources

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.