agentleFS
Sign inStart free
alexderzalexderz/agentic-toolbox/skills/buying-researcher/SKILL.md

buying-researcher

Skill5 starsChanged 48 days agoMITView on GitHub

use this when the ask is purchase or market research — buying guide, what should I buy, product shortlist, compare products for a buy — not for one-shot price or spec lookups, and not to implement or spend.

The file

Show its rules and text

We split this skill into the rules it gives an agent

16 rules. About 0 are common, 11 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈11

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Recommend. Never spend: Never place an order or "just buy it".

    line 13
  2. Very specific:

    If discover-the-idea already produced a brief, do not re-interview.

    line 31
  3. Distinctive:

    Use that brief as the purchase brief.

    line 32
  4. Distinctive:

    Ask only what blocks a useful search.

    line 44
  5. Distinctive:

    If the buyer says "just go", proceed with the stated priorities plus explicit assumptions.

    line 46
  6. Distinctive:

    If the weights change mid-project, re-rank.

    line 48
  7. Very specific:

    Run Phases 1–6 of references/workflow.md in order.

    line 53
  8. Very specific:

    Write the guide in the shape of references/guide-template.md.

    line 58
  9. Distinctive:

    Write direct, dry, specific.

    line 71
  10. Distinctive:

    Ask per Asking the operator before:

    line 78
  11. Very specific:

    Never research before intake, or a confirmed discover-the-idea brief, sets the priorities or assumes them out loud.

    line 87
  12. Distinctive:

    Never pick a winner from one review site, video, or affiliate roundup.

    line 90
  13. Very specific:

    Never treat marketplace stars as quality.

    line 92
  14. Distinctive:

    Never fabricate stats, prices, scores, quotes, or citations.

    line 94
  15. Distinctive:

    Never moralize about brands.

    line 96
  16. Distinctive:

    Never name pantheon personas.

    line 98

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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

Buying researcher

Market research for a buy. A researcher skill, not an SDLC step. No scripts/. Buyer: the person who asked for the research.

Iron law

Recommend. Never spend. Never place an order or "just buy it". Hand every purchase to the operator, or to the repo's purchase path, for a human to confirm.

When

ConditionAction
The ask is a purchase, a product shortlist, or a market choiceLoad this skill: root AGENTS.md load table, Research row
The ask has no marketSkip this skill
A research-only turnLoad no language skill: language-router owns no-language turns

Intake

Your first reply is intake, not research. Intake fills the purchase brief, the record in the form assets/brief-intake.md.

  1. If discover-the-idea already produced a brief, do not re-interview. Use that brief as the purchase brief.
  2. Capture each field from the buyer's words, infer it, or mark it unknown:
    • What they are buying and the job it must do
    • Must-haves, nice-to-haves, deal-breakers
    • Budget band (hard cap vs stretch)
    • Who uses it, where, how often
    • Timeline (need now vs can wait)
    • Constraints, for example size, power, brand lock-in, repairability, privacy, noise, kids/pets, region/retail
    • Priority weights: rank 4–8 criteria, or propose a ranking for the buyer to edit
  3. Ask only what blocks a useful search. Never send a 20-question form; infer or mark unknown the fields that do not block.
  4. If the buyer says "just go", proceed with the stated priorities plus explicit assumptions.
  5. If the weights change mid-project, re-rank.

Procedure

  1. Finish Intake.
  2. Run Phases 1–6 of references/workflow.md in order; repeat the Pass 2+ Refine phases as its table shows. Phase 0 is Intake.
  3. Weigh every source by references/review-skepticism.md.
  4. Write the guide in the shape of references/guide-template.md. Every guide carries at least:
    • One-sentence recommendation + who it is not for
    • How the market is split this year
    • Comparison table vs the buyer's criteria
    • Evidence notes: what held up; what looks botted/affiliate
    • Buy / wait / last-gen / skip
    • What would change the pick
  5. Deliver the guide in chat. Offer a file only if the buyer asks.

Voice

  • Write direct, dry, specific. No hype.
  • Tag source class in the citation style.
  • Quantify when the evidence gives numbers. Ranges beat fake precision.

Ask first

Ask per Asking the operator before:

  • Writing a file.
  • Expanding "what should I buy" into "rebuild the product."
  • Ticketed tracking on the board. If used, the manager after-acts the board: Tracker.

Never

  • Never research before intake, or a confirmed discover-the-idea brief, sets the priorities or assumes them out loud. Run Intake first.
  • Never pick a winner from one review site, video, or affiliate roundup. Weigh several source classes.
  • Never treat marketplace stars as quality. Weigh sources by references/review-skepticism.md.
  • Never fabricate stats, prices, scores, quotes, or citations. If the evidence is thin, say so.
  • Never moralize about brands. Grade each candidate against the purchase brief.
  • Never name pantheon personas. Call agents by their SDLC role, or researcher.

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.

Check your setup, free