outbound-strategy
Installation
SKILL.md
An agent that turns a single domain into the targeting half of an outbound strategy: who to go after, why they buy, and which intent signals reach them while they are in-market. It stops at foundations plus signals — messaging and campaigns are separate steps.
Input
website_url— the company the strategy is for.- (optional)
market_geo,acv_range,sales_motion(PLG / sales-led / hybrid),icp_constraints.
Procedure
- What they sell — category, top use cases, who it is for, pricing and motion clues. Prefer the site, docs, pricing, case studies, reviews.
- Who buys — economic buyer vs. daily user; which teams feel the pain; industries named or implied.
- Competitors and alternatives — direct rivals, plus the "do nothing / build in-house / spreadsheet / incumbent" alternatives, and the switching trigger.
- Segment into ICPs by shared buying reason, not by firmographic label. A segment is real when its members would buy for the same why. Bound each one — company type, industries, employee-size bands, geographies — tightly enough that two people would build the same list from it. Rank by fit × reachability: a perfect-fit segment you cannot find at volume is a worse ICP than a good-fit one you can.
- Draw personas, one title to one persona. Assign titles in the person's own language, not the company's org-chart language. A given title belongs to exactly one persona — if "Head of Growth" could sit in two, the personas are wrong; redraw them. Keep economic buyer, daily user, and champion distinct, and state what each cares about (buyer: outcome/ROI; user: friction; champion: looking good internally).
- Map value props to pains. Take the pains in the buyer's words, and for each write the value prop that resolves it — the outcome the buyer gets, phrased as a change in their world, not a feature name. A value prop with no pain behind it is a feature: drop it or find its pain. Where a metric, customer, or result backs a claim, cite it; where none exists, mark it unproven rather than dressing it up.
- Shortlist signals — 3-5 detectable signals, each with a why-now.
Label anything you cannot verify from public sources as an Assumption and keep it plausible.