sales-research

Files (8)

• README.md
• SKILL.md
• openai.yaml · agents/
• README.md · evals/
• evals.json · evals/
• practitioner-routing.md · references/
• selling-is-leading.md · references/
• sources.md · references/

SKILL.md (first lines)

# Sales Research

Turn the available evidence into a concrete sales decision or next step. Match
practitioners to the situation and let the sales interaction demonstrate how
the engagement would work.

## Classify the situation

Identify the mode, stage, buyer, decision, and main risk:

- **Mode:** bespoke consulting, SaaS or product, employment or recruiting,
  partnership or community or mentoring, or existing-client delivery.
- **Stage:** positioning or outreach, qualification, discovery, paid diagnostic,
  proof or POC, proposal, pricing or negotiation, delivery or expansion, or
  late-stage indecision.
- **Buyer:** founder or small team, sophisticated AI user, enterprise, civic or
  government, connector, or end user.
- **Decision:** what this answer or interaction should resolve.
- **Risk:** premature advice, free diagnosis, broad positioning, jargon,
  self-discounting, weak qualification, or a vague next step.

Ask for the smallest missing fact only when it would materially change the
recommendation. Otherwise state the assumption and proceed. Apply consulting
free-versus-paid boundaries contextually; interviews, mentoring, partnerships,
product support, and already-paid delivery have different expectations.

## Gather evidence and select practitioners

Start with the user's supplied facts, relevant project instructions, prior
conversations, sales playbooks, and existing research when available. Retrieve
only context relevant to this decision. A fresh installation needs no private
corpus, other skill, account integration, or project-specific directory.

Read [practitioner-routing.md](references/practitioner-routing.md). Choose one
to three practitioners whose work best matches the situation; use four only
when materially necessary. If the user has a Sales Practitioner Registry,
read the selected entries' tags, best use, and limitations. Missing fields
remain unknown.

Use [sources.md](references/sources.md) to locate the core primary publishers
and podcasts. The routing list selects candidates; it does not substantiate
claims. Attribute a practitioner claim only after retrieving its primary source
or checking a direct citation already in the user's project. Keep links to the
specific article, episode, or book passage supporting the claim.

Search outward when local evidence lacks a comparable case, a material claim
may be stale, credible sources disagree, or the user requests broader research.
Prefer primary work and detailed firsthand cases from people who sold comparable
software, AI, or professional services. Label self-reported results and inference.
Reject generic SEO pages, AI roundups, unsupported viral advice, and manipulative
pressure tactics. Use technical or procurement primary sources when delivery
constraints change the sale.

If browsing or a source is unavailable, explain that limit and work from the
supplied evidence. Keep unsupported practitioner attributions out of the answer;
do not invent buyer facts, results, prices, quotes, or citations.

Resolve conflicts in this order:

1. safety, policy, and project instructions;
2. the explicit request and current facts;
3. the user's current constraints;
4. evidence from the closest comparable situation;
5. practitioner primary sources;
6. general frameworks and community opinion.

State a disagreement when it changes the recommendation. Stop researching once
the decision is adequately supported.

## Recommend or draft

For prospect or client communication, read
[selling-is-leading.md](references/selling-is-leading.md). Give the copy first,
with a clear decision or next step. Keep practitioner names and framework
analysis internal unless the user asks to see the research.

For advice or research, lead with the recommendation, then the evidence that
changes the decision, material assumptions or disagreements, and the next
action with an owner. Distinguish observations, source claims, and inference.

For a call review, ground feedback in the supplied transcript. Separate effective
behaviors, missed opportunities, and the next-call changes worth practicing.

A narrow drafting question needs a usable answer, not an unsolicited research
memo. When the user requests a saved memo, follow their project conventions and
include the date, scope, recommendation, uncertainty, and direct source links.
Preserve private buyer information and negotiation details in their authorized
context. Publishing or sending a draft requires separate authority.

Finish when the answer supports the decision, or identify the exact missing fact
when progress depends on information only the user can provide. Do not send
messages, publish, deploy, or change external systems without authorization.

From ThatGuySam/skills, MIT.

More from ThatGuySam/skills