prospect-research

name	prospect-research
description	Build or refresh a Research Capsule for a sales prospect. Pulls background, workflow pain points, recent signals, and three outreach angles (V1/V2/V3). Writes a structured Markdown file to your local vault. Read-only on social sources — never sends or posts anything.
version	1.0.0
author	Bubble Invest
license	MIT
allowed-tools	["Read","Write","Edit","Bash","WebSearch","WebFetch"]

prospect-research

Build or refresh a Research Capsule for a sales prospect: a structured Markdown file in your local vault plus an optional short summary projection you can paste into your CRM.

When to invoke

• On demand: "Research [prospect name / company]" or "refresh capsule for [slug]"
• In a pipeline: call this skill after a discovery step produces a new lead row, before drafting outreach
• Batch: iterate over a list of slugs sorted by next_refresh_at ASC, research each one before handing off to your outreach step

Inputs

• slug (required) — a short kebab-case identifier for this prospect (e.g. jane-smith-acme). You define the slug convention; keep it stable across refreshes.
• force_refresh (optional, default false) — if true, recompute all sections even when a capsule file already exists. If false and the file exists, append a refresh timestamp line to the History section and stop.

Configuration — fill in before first use

Before running this skill, set your context in prospect-research/config.md (create it from the template at prospect-research/templates/config-template.md):

• ICP definition — who your ideal customer is (industry, role, company size, geography, problem type)
• Your product/service — what you offer and the core value proposition
• Outreach angles — customize the V1/V2/V3 framework (see below) to fit your offer
• Vault directory — where Research Capsules will be written (~/your-vault/Capsules/ or any absolute path)
• CRM field name — the label for the short summary projection in your CRM (optional)

Process

1. Check for an existing capsule at {vault_dir}/Capsules/{slug}.md.

   • If it exists and force_refresh=false → append a one-line refresh note to the History section and stop.
   • Otherwise → full research.

2. Gather prospect data — collect the following. If a source is unavailable or uncertain, write To enrich. (do not invent).

   • Current role and background: LinkedIn profile, company website, press releases. Note title, company, tenure, and any recent role change.
   • Company context: industry, size, business model. Public filings, Crunchbase, or company site.
   • Mutual connections / existing relationship: note any shared contacts, past interactions, or referrals.
   • Recent signals: LinkedIn activity (posts, comments), job postings at their company, recent news (funding, product launch, M&A, regulatory event). These are the hooks for V3.
   • Workflow pain: infer from their role + industry + your ICP definition. What time-consuming, error-prone, or high-stakes workflow does your product address for this person? Be specific (e.g. "manual reconciliation of X", "copy-paste between Y and Z").

3. Build three outreach angles — adapt these to your product using your config.md:

   • V1 — Workflow pain: the concrete, manual task your product removes. Frame it from their daily reality, not your feature list.
   • V2 — Expertise / peer credibility: a "practitioner-to-practitioner" angle — you understand their domain, not just their job title.
   • V3 — External signal: a rebound off a recent market event, regulatory change, industry consolidation, or news item relevant to them. Shows you did your homework.
   • Recommended angle: pick V1/V2/V3 based on what you found. Justify in one sentence.

4. Recommended action: propose a next step (draft-outreach, warm-engage, research-more, or skip). Base it on fit with your ICP + strength of signals found.

5. Render the capsule using lib/capsule.py:render_capsule(research_input) and the template at templates/capsule-template.md.

6. Write the capsule file to {vault_dir}/Capsules/{slug}.md (idempotent overwrite).

7. Compute the CRM projection (optional): lib/capsule.py:compute_notes_projection(research_input) → a ≤ 200-character summary suitable for a CRM notes field.

8. Log a line in {vault_dir}/Daily/{YYYY-MM-DD}.md: skill name, slug, recommended angle, proposed action, capsule path.

Output

• Capsule path — absolute path of the written .md file
• One-line summary for your log or CRM:

  {slug} — capsule {created|refreshed} — angle V{1|2|3} — action {draft-outreach|warm-engage|research-more|skip}

Guardrails

• Read-only on social sources. This skill never posts, likes, comments, or sends messages. Outreach is a separate, human-approved step.
• Do not invent signals. If a piece of information cannot be found or verified, write To enrich. — never fabricate data to fill a section.
• Defensive CRM writes. If you integrate CRM writes, wrap them so a transient API error never prevents the capsule file from being written (the .md is the primary output).
• Idempotent. Re-running without force_refresh=true only adds a history line — nothing is overwritten.

Files in this skill

prospect-research/
  SKILL.md                      # this file — skill instructions for Claude
  lib/
    capsule.py                  # pure-Python rendering helpers (stdlib only)
  templates/
    capsule-template.md         # Markdown template for Research Capsules
    config-template.md          # Fill this in with your ICP + vault path
  examples/
    example-research-input.py   # Annotated example of a research_input dict

Quick start

1. Copy templates/config-template.md to prospect-research/config.md and fill it in.
2. Ask Claude: "Research [Name] at [Company] — slug: name-company".
3. The capsule lands in your configured vault directory.

Python helpers

See lib/capsule.py for the public API:

• render_capsule(research_input, template_path=None) -> str
• write_capsule_file(slug, content, vault_dir) -> Path
• compute_notes_projection(research_input, max_len=200) -> str
• append_interaction_line(capsule_path, line) -> None

All functions are pure Python (stdlib only), fully documented, and tested independently of any CRM or database backend.