agentleFS
vcxcviivcxcvii/rainmaker/skills/know-my-buyer/SKILL.md

know-my-buyer

Skill0 starsChanged 4 months agoMITView on GitHub

Interview the site owner about who buys, in their own words, grounded in a real diagnosis rather than generic questions, and write the shared business context every other skill reads. Refuses to run without a diagnosis.

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 2 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Run rainmaker context --check first.

    line 23
  2. Very specific:

    Only the data/ files listed in this skill's Consumes table.

    line 31
  3. Very specific:

    If strategy.json.context_hash does not match the current hash of context/business.md, say exactly:

    line 33
  4. Very specific:

    If context/business.md carries confidence: stub, continue, and stamp every output with: "Built on a stub context.

    line 39
  5. Very specific:

    Read diagnosis.json. Open with three sentences of fact and nothing else: the tier distribution, the top 3 findings by revenue_score, and …

    line 61
  6. Distinctive:

    Ask one question at a time.

    line 62
  7. Very specific:

    Never invent a buyer_language quote.

    line 73
  8. Distinctive:

    Never write a proof point without a source.

    line 74

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

know-my-buyer

Context load

Run rainmaker context --check first. It prints what exists, what is stale, and exits 1 if anything this skill requires is missing.

Then read, in this order:

  1. context/business.md in full. If absent, stop: "No business context. Run rainmaker audit, then the know-my-buyer skill."
  2. context/voice.md if this skill writes prose. If absent, stop and say so.
  3. context/glossary.md if this skill names products, features or competitors.
  4. data/strategy.json if this skill reads or writes strategy.
  5. Only the data/ files listed in this skill's Consumes table. Never crawl or call an API the core already covers.

If strategy.json.context_hash does not match the current hash of context/business.md, say exactly:

"Business context was edited after the strategy was written. Re-run know-my-buyer, or run rainmaker context --sync to accept the prose as authoritative."

Then stop.

If context/business.md carries confidence: stub, continue, and stamp every output with: "Built on a stub context. Nothing in it came from a buyer. Run know-my-buyer to replace it."

Override for this skill only: a stub is exactly what know-my-buyer exists to replace. Do not stop, and do not repeat the stamp above during the interview itself; the stamp is for skills that read the context, not the one that writes it.

Consumes

FileWhyIf missing
data/snapshots/<latest>/diagnosis.jsontier distribution and top findings to open withrefuse, see below
data/competitors.jsoncontrast points for the opening summarycontinue without it, note the gap
context/business.md, data/strategy.jsonthe stub or prior interview to build onrainmaker context --init first if truly nothing exists

Produces

context/business.md (new version, prior archived to context/history/), data/strategy.json (version bump, prior archived to data/strategy-history/).

Refuses when

No diagnosis exists. Print exactly: "Run rainmaker audit first. Interrogating you about a site I have not looked at produces generic questions." Do not proceed under any circumstance, including a user insisting they already know their answers: the value of this skill is that every question cites a number, and there is no number to cite yet.

Procedure

  1. Read diagnosis.json. Open with three sentences of fact and nothing else: the tier distribution, the top 3 findings by revenue_score, and the sharpest contrast against competitors.json if it exists. No adjectives, no framing, just the numbers.
  2. Ask one question at a time. Never batch, never parallelise. Question N depends on the answer to N-1, and that dependency is the entire value of an interview over a form.
  3. Ask a minimum of 12 questions, covering, in this rough order: who pays, who blocks the deal, what they were doing before finding us, the exact words used in the last deal won, the exact words used in the last deal lost, which page sales actually sends before a close, which page the diagnosis says is broken that sales still sends anyway, which competitor came up in the last three deals, what we refuse to claim and why, what we would bet the quarter on, what would prove this whole strategy wrong.
  4. Every question must cite a specific finding, number or competitor fact from the diagnosis. A question that could be asked of any company is a defect. Bad: "Who is your target customer?" Good: "Your /pricing page carries 1,240 impressions and 11 clicks over 28 days, sitting at position 8.4. Who reads that page before a deal closes, and what do they still not know when they leave it?"
  5. When an answer is generic ("enterprise companies", "anyone who needs X"), push back once, quoting their own words back at them, and ask for the specific deal it happened in. Accept the second answer even if it is still imperfect; two rounds is the budget, not zero and not five.
  6. As pain points emerge, write them immediately with an id, the verbatim buyer_language, and the evidence type. A pain point with an empty buyer_language array is a defect. If the interview has not yet produced a direct quote for a pain point you believe exists, ask again rather than paraphrasing one into existence.
  7. At the close, write both artifacts via rainmaker context --sync semantics: update context/business.md in full, matching every persona, pain point, proof point and competitor to the same id in data/strategy.json. Bump strategy_version. Record a decisions entry for every field that changed from the prior version, with a one-line reason.
  8. Summarise what changed: which prior beliefs were confirmed, which were overturned, and which questions remain open (write these to the "Open questions" section of business.md rather than leaving them unrecorded).

Decision rules

  • 12 questions is a floor, not a target to race past. A fast-moving user is not a reason to cut the pain-point or objection lines of inquiry short.
  • Never invent a buyer_language quote. Paraphrase is a downgrade and must be marked as such if used at all (status: hypothesis, never validated).
  • Never write a proof point without a source. "They said their close rate improved" without a proof id is not evidence.
  • The "what we will not say" section is mandatory even when the user has nothing to add. Ask once directly: "Is there any claim you would not want us to make, even if it were technically true?" If the answer is genuinely nothing, write that the section was asked and nothing was named, rather than leaving it blank with no record it was asked.

Output

The live interview is conversational, one question visible at a time. The closing summary:

## Business context updated: version <n>

Confirmed: <beliefs that held>
Changed: <beliefs that moved, and why>
New: <personas, pain points, proof added>
Open questions: <what the interview could not settle>

context/business.md and data/strategy.json are in sync (hash <short hash>).

Done when

At least 12 questions were asked one at a time, every pain point has non-empty buyer_language or is explicitly marked hypothesis, both artifacts are written with matching ids and a matching context_hash, the prior version is archived, and every changed field has a decisions entry with a reason.

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.