Battlecard

Instructions

State check. This skill assumes the org has at least basic positioning established — a value prop and competition notes saved to org knowledge. Load org competition knowledge via swan-get-memory. If there's no competition slot at all and no value prop on file, the battlecard will end up generic — build out positioning first (the positioning capability has its own setup flow) before generating a battlecard. If competition notes exist for some competitors but not this one, that's fine: this run becomes the create-from-scratch path.

When to use

• "Build me a battlecard for X."
• "Update our notes on competitor X."
• "How do we beat X?"
• Before a competitive deal cycle.

What it produces

A six-section battlecard in canonical format, saved to org knowledge via swan-update-knowledge-competition and returned to the user.

---

Step 1 — Identify the competitor

Get the competitor name and primary URL from the user. Confirm the spelling. If the user has prior notes, load them first via swan-update-knowledge-competition (the read path is via swan-get-memory on the competition slot).

Step 2 — Pull existing intelligence

swan-get-memory for org competition knowledge. If there's already a section on this competitor, treat the task as update, not create from scratch. Preserve what's true; only revise what's stale.

Step 3 — Gather fresh signals

Make a small number of cheap queries. Do not dump everything; pick the highest-signal sources first.

• swan-fetch-scraped-url on the competitor's homepage and pricing page. Capture: the headline value prop, the primary CTA, the named target customer/segment.
• swan-fetch-scraped-url on their /customers or case-study page if linked. Capture 3-5 named customers and any pattern.
• swan-linkedin-social-media-presence on the competitor's company LinkedIn. Capture: posting cadence, hiring areas (titles in recent posts), leadership announcements.
• swan-website-traffic on the competitor's domain. Capture: traffic order of magnitude (10k, 100k, 1M monthly) and trend direction.
• swan-fetch-business-events for funding/leadership/partnership events.

If the user already has fresh intelligence, skip the corresponding fetch. Don't burn credits re-pulling.

Step 4 — Compose the six sections

Write the battlecard in this exact structure:

## Battlecard — 
*Last updated: *

### 1. One-line description

### 2. Where they win

### 3. Where they lose

### 4. Common objections we hear

### 5. Win conditions

### 6. Where to push in a deal

Be honest about strengths. A battlecard that pretends the competitor has no advantages doesn't get used.

Step 5 — Save the result

swan-update-knowledge-competition with the new battlecard content. Use append or find-and-replace depending on whether this is a new competitor or an update.

Step 6 — Return to the user

Return the full battlecard in the response. The user typically wants to read it once and share it.

---

Rules

• MUST cite real evidence in strengths and weaknesses sections — not opinion. If you can't cite, leave it out.
• MUST keep total length under one page of dense text (~600 words).
• MUST save the final to swan-update-knowledge-competition. A battlecard that lives only in chat history is wasted.
• NEVER include speculation framed as fact. If a weakness is rumored, label it ("reported in G2 reviews").
• NEVER include legally sensitive claims (e.g. "their product is insecure") without a citation.
• If the competitor has < 1,000 monthly visitors and no LinkedIn presence, flag that — usually means the user is fighting a ghost. Suggest they reassess whether this competitor is worth a battlecard.
• Date-stamp the battlecard. Competitive intel rots fast.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: swan-gtm
• Source: swan-gtm/gtm-skills
• License: MIT

Install and usage instructions live in the source repository linked above.