Competitive Intel

Turn research on one competitor into three sales documents that agree with each other. A plugin for Claude — works for any company, in any industry.

Most competitive intelligence dies in one of two ways. Either it's a 15-page document nobody opens before a call, or it's a battlecard with confident claims nobody can source — and the first time a prospect says "actually, we shipped that last quarter", the whole thing loses its credibility.

This builds all three artifacts from one research pass, so the long one carries the evidence, the short one carries the conclusions, and the page a seller holds during the call is generated from the card rather than written separately. They cannot drift apart and contradict each other in front of a customer.

Document	Length	Who reads it, and when
Deep-dive	9–15 pages, .docx	Product marketing, strategy, RFP responses. Every claim dated, sourced, linked
Battlecard	2–3 pages, .docx	A seller, five minutes before the call
One-pager	1 page, PDF	A seller, during the call, on paper

_{Worked examples, built by the plugin. The vendor, reviews and numbers are invented — see docs/examples for the full PDFs.}

Install

Option 1 — the packaged plugin. Download dist/competitive-intel.plugin (also attached to Releases) and open it in Claude. Accept the install card. Nothing else to set up.

Option 2 — Claude Code. This repository is also a marketplace, so:

claude plugin marketplace add kaordones/competitive-intel-plugin
claude plugin install competitive-intel@kaordones-plugins

Either way, install the document dependencies once:

cd <plugin>/shared/scripts && npm install

Quick start

New here? Getting started is a twenty-minute walkthrough — check your setup against the shipped example, fill in your profile, then produce a real set of three documents.

There is one setup step, because every document compares a competitor against you:

"Set up my competitive profile"

Claude interviews you — what you ship (and what's still roadmap), who buys it, your value-promise axis, your defensible differentiators, where you honestly lose, pricing, and where your internal evidence lives — then writes competitive-profile.md in your working folder. Fill it once; share the file with your team.

Then:

"Do a deep dive on <competitor>"           → the sourced reference document
"Build a battlecard for <competitor>"      → the 2–3 page sales card
"Make one-pagers for all our competitors"  → one combined, badged PDF

For a competitor you've never researched, go deep-dive → battlecard → one-pager. The research forces the precision that makes the card sharp. A standalone card is a perfectly normal request too.

The part that makes it trustworthy

Documents are written as structured JSON and rendered by a shared script, so every card in your set looks identical and your effort goes into argument instead of formatting. But the build also reads what you wrote and flags:

• Unhedged denials — "they don't have X" where your research only supports "no public evidence of X". This is the claim that loses a deal when the feature quietly shipped last quarter. Scoped to claims about the competitor, so honest concessions about your own product aren't flagged.
• Judgement words — opinions a prospect can argue with, and the sentences you'd least like to see forwarded.
• Undated quotes — a review quote with no date can't be re-checked, which defeats the point of sourcing it.
• Missing snapshot rows the one-pager will need later, caught while the card is still open rather than a week later.
• Length against the format's budget, with the heaviest blocks named, so you trim before rendering rather than after.

None of them block the build. They're the review a careful colleague would give.

The rules the documents are built on

1. Sourced or hedged — never confident and unsourced.
2. Date every external fact. The same complaint seven years apart is a finding, not an anecdote.
3. Attack structure, not quality. "There is no approval workflow, and their own reviewers say so" ends an argument; "their reporting is bad" starts one.
4. Concede real strengths out loud. It buys the right to be believed everywhere else.
5. Include honest walk-away signals. Sales trusts a card that tells them when to stop.
6. Write as though the competitor will read it — because eventually one does.

What's inside

skills/
  competitive-profile/     the setup interview → competitive-profile.md
  competitor-deep-dive/    long sourced document + research playbook
  competitor-battlecard/   the 2–3 page card
  competitor-one-pager/    the single A4 page, generated from a card
shared/
  references/              content schema, company-profile template
  scripts/                 renderer, .docx build + checks, one-pager PDF build
  assets/                  worked examples (entirely fictional)

Each skill loads on its own when your request matches it — you don't call them by name. See CONNECTORS.md for the tool categories (chat, knowledge base, CRM, meeting notes) the research steps can draw on; none is required.

Requirements

• Node with the docx package, for the .docx documents
• LibreOffice (soffice), to render .docx → PDF for proofreading
• Headless Chromium, for the one-pager PDF — the A4 layout is tuned to Chromium's print engine, and other converters won't reproduce it

A note on the examples

The worked examples use an invented vendor, invented reviews and invented numbers, purely to show the tone and density that make these documents work. Nothing in them is a claim about any real company. Replace every word.

License

MIT — see LICENSE. Use it, change it, sell what you make with it.