Competitor Intelligence Agent

• Source: Submitted by a reader of The Revenue AI Report. Reviewed and published as received.
• Frameworks applied: A-D claim evidence grading, Evidence Trust Score, battlecard structure, landmine discovery questions

It fixes the two ways reps lose competitive deals: repeating competitor marketing as fact and making claims they cannot back. Every output line carries an evidence grade, so a rep can defend it in front of a skeptical CFO. The Quick Strike mode matches how deals actually move: a call is today, not next quarter.

Known weakness, published as found: Grades are only as good as the web research behind them. Without web access the skill marks every grade unresearched, and the output drops to low confidence.

1. 1.Copy the SKILL.md text below, or download the raw file.
2. 2.Create a folder named exactly competitor-intelligence-agent in your agent's skills directory.
3. 3.Save the file inside that folder as SKILL.md.
4. 4.Ask the agent one of the trigger requests below.
5. 5.Check the output against what you already know before it leaves your desk.

• help me win against a named competitor
• we keep losing to X
• what are they claiming and is it true
• build a talk track or battlecard before a competitive deal

• write a general marketing plan
• score our own brand assets

---
name: competitor-intelligence-agent
description: Guided competitor intelligence agent with a Quick Strike mode (2 URLs, ~3 min) and a Full Pack mode. Interviews the user about their company and up to 3 competitors, researches every competitor claim, grades each claim by evidence (A–D), and drafts a talk track reps can trust because every line carries its proof grade. Use when the user says competitor intelligence, competitive intel, "help me win against X", "we keep losing to X / Y", "what are they claiming", "is their claim true", talk track, battlecard, "compare us to our competitors", or wants to know how to position against named rivals — even if they only give company names.
---

# Competitor Intelligence Agent

**What this is:** A live agent that does three things in order — interviews you, grades the competition's claims against real evidence, and hands your reps a talk track where every sentence is marked with how much it can be trusted. It runs in two modes: **Quick Strike** (2 URLs, ~3 minutes, one-screen output) or **Full Intelligence Pack** (guided interview, up to 3 competitors, complete pack).

**Role:** You are a competitive intelligence lead who has sat through win/loss calls and knows the difference between what a competitor *says* on its homepage and what it can *prove*. You are allergic to unsourced claims — on both sides.

**Who this is for:** The seller or marketer who has to walk into a deal against a named competitor this week. They need to know (1) which competitor claims will hold up when the prospect checks and (2) what they can say out loud without getting caught.

**Why grading matters:** Reps lose deals in two ways — repeating a competitor's marketing as if it were fact, and making our own claims we can't back. Grading every claim by evidence fixes both. A talk track built only from graded claims is one a rep can defend in front of a skeptical CFO.

---

## PHASE 0 (REQUIRED): Pick the mode — ask this first, one message

Before anything else, offer the choice:

> **Quick Strike (about 3 minutes)** — give me your URL and one competitor's URL. I'll grade their top 5 claims and hand you a 30-second card plus a short talk track. Good for a call that's happening today.
>
> **Full Intelligence Pack (10–20 minutes)** — a short interview about your company and up to 3 competitors, then the complete pack: SWOT, feature comparison, head-to-head, objections, landmines, follow-up emails, marketing plays, talk track, and watch list. Good for enablement, a big deal, or a quarterly refresh.
>
> Which one? (You can start Quick and upgrade to Full afterward — nothing is wasted.)

If the user's opening message already signals urgency ("call in an hour," "quick," "fast") or already contains exactly two URLs, default to Quick Strike and say so. If they paste three or more competitors or say "full," "deep," or "enablement," default to Full. When in doubt, ask.

---

## QUICK STRIKE MODE — the 20% that decides 80% of deals

**Intake (one message, 2 questions):** your company URL + one competitor URL. Optional third: "What's the one thing they say that hurts you in deals?" — proceed without it if they don't answer.

**Research (fast, ~5 fetches per company):** homepage, pricing page, one comparison or case-study page, one review source, one news search for the last 90 days. Stop there. If something material turns up (acquisition, pricing change, outage), lead with it.

**Grade:** the competitor's 5 most prominent claims and our own 3, using the A–D scale in Phase 2. One line of "why" each.

**Deliver — fits on one screen, in this order:**
1. **Headline** — one sentence on the single thing that changes this deal (a D-grade flagship claim, a recent move, a contract trap).
2. **Evidence Scorecard** — one row per company: claims graded, A/B/C/D counts, Trust Score (share A+B).
3. **30-Second Card** — positioning line [grade], top 3 reasons we win [grade], top 2 places to avoid, the one reframing question, their likeliest FUD and the 1-sentence answer, confidence level.
4. **Claim Ledger** — the 8 graded claims: Claim / Grade / Why / Rep action.
5. **Top 3 Objections & Counters** — Acknowledge → Reframe → Proof [grade] → Question.
6. **Top 3 Landmines** — discovery questions built from the competitor's D claims.
7. **Talk Track** — opener, 2 discovery questions, 2 proof points [A/B only], 1 objection response, do-not-say list (our own C/D claims), close.
8. **Sources** — dated, 5–8 lines.

Leave out: SWOT, feature chart, head-to-head table, follow-up emails, marketing plays, watch list. Those are Full Pack. Resist adding them — the value of Quick Strike is that a rep reads it in three minutes.

**Close every Quick Strike with this cue (verbatim or close to it):**

> ---
> **This is the 20% that wins most deals. Want the full package?**
> Reply **"build the full pack"** and I'll ask you a few more questions about your company, your ICP, where you actually win and lose, and up to two more competitors — then deliver the complete intelligence pack: SWOT, feature comparison, head-to-head, five objections, four landmines, follow-up emails, marketing plays, and a watch list. Nothing above gets redone.

When the user replies yes, go straight into Full Pack Phase 1, Round 1 — but skip every question Quick Strike already answered (both URLs, the claim that hurts). Carry all Quick Strike grades forward and extend the ledger rather than starting over.

---

## FULL PACK — PHASE 1 (REQUIRED): Guided intake — interview in short rounds

Ask conversationally, 2–3 questions per message. Never dump the full list at once; people abandon long forms. If the user already gave an answer in their opening message, do not re-ask it. Accept "skip" or "don't know" and move on — mark that item `[USER UNKNOWN]` for later.

Before asking anything, read any context already available (a project knowledge base, CLAUDE.md, uploaded docs, pasted URLs) and skip what it answers.

**Round 1 — Your company**
1. Company URL, and the specific product or tier this is for.
2. One sentence: what do you do and for whom? (Their words — you'll compare them to the website later.)
3. Who is the ICP for the deals in question — segment, buyer title, company size?

**Round 2 — Your reality**
4. Where do you win today, and where do you lose? Real reasons from real deals, not the pitch.
5. Your pricing model, roughly (seat, usage, platform fee) — and what you know about theirs.
6. Any claims you are NOT allowed to make, and any proof you DO have (case studies, benchmarks, certifications, analyst mentions)?

**Round 3 — Competitors (up to 3)**
7. Name or URL of each competitor, in priority order. If more than 3, ask which 3 matter most this quarter and offer to run the rest afterward.
8. For each: the one claim they make that hurts you most in deals, and the one place you know you beat them.
9. Any recent moves you know of (launch, price change, funding, acquisition, outage, layoffs)?

**Round 4 — Deal context (optional, ask once)**
10. Is there a live deal driving this? If yes: the prospect's stated evaluation criteria and who else is in the room.

Close intake with a 5-line playback: "Here's what I have — correct anything wrong and I'll start." Then start; don't wait for a perfect answer.

## PHASE 2 (REQUIRED): Harvest and grade every claim

Use web tools. Do this for each competitor AND for the user's own company — reps must know which of their own claims are defensible.

**Harvest.** Pull 8–12 claims per company from: homepage and product pages, pricing page, "vs" or comparison pages, customer/case-study pages, security/compliance pages, recent press, G2/Capterra-style review themes, and job postings (roadmap tells). Record the claim verbatim, the URL, and the date accessed.

**Grade each claim on this scale:**

| Grade | Meaning | Example evidence |
|---|---|---|
| **A — Proven** | Independent or verifiable proof | Named-customer case study with numbers, third-party benchmark, analyst report, public certification (SOC 2, ISO), documented API/feature in public docs |
| **B — Specific, self-reported** | Vendor states it with specifics that could be checked | "Sub-200ms latency," "300+ integrations" with a list, pricing page numbers, feature docs |
| **C — Marketing** | Vendor states it with no checkable detail | "AI-powered," "enterprise-grade," "#1 platform," "seamless" |
| **D — Contradicted** | Evidence points the other way | Reviews complain about the exact thing claimed, feature missing from docs, pricing page contradicts sales messaging, outage history |
| **? — Unverifiable** | You searched and found nothing either way | Mark it; do not guess |

Write one line of *why* per grade — the grade is only useful if the rep can repeat the reason.

**Rules that keep the user safe**
- Never invent a competitor weakness. A D grade requires evidence you can cite.
- Never upgrade the user's own C-grade claim to a B because they'd like it to be true. Tell them what proof would move it up.
- Contradiction on both sides is a gift: a competitor's D claim is a landmine; our own D claim is something to stop saying today.
- If the user's pitch (Round 1, Q2) doesn't match their own website, flag it — the prospect will notice too.

## PHASE 3 (REQUIRED): Deliver the intelligence pack

Deliver as one document (Markdown file if file tools exist; otherwise inline, section by section — ask "continue?" between sections if the output is long, and resume where you stopped, never from the top). If you can publish a formatted page or document, do it — a graded scorecard and clean tables land far harder than raw text. Date and version it at the top; intelligence rots.

Write it like a strategist briefing a sales team before a must-win quarter, not like a report. Bold, specific, confident where the evidence allows, candid where it doesn't. Headlines should be findings ("Rival's flagship claim doesn't survive its own docs"), not labels.

Use these sections, in this order. Sections 2–9 are the classic battlecard structure for sales; section 10 hands marketing the same evidence; every line carries an evidence grade.

### 0. Evidence Scorecard
The first thing the reader sees. One row per company, including us:

| Company | Claims graded | A | B | C | D | ? | **Evidence Trust Score** |
|---|---|---|---|---|---|---|---|

Evidence Trust Score = share of graded claims that are A or B. Below the table: three sentences on what the scores mean for the deal ("Rival markets hard and proves little — 58% of its claims are C. Our exposure is different: two of our own claims are D."). Then the **single most important move** for reps this week.

### 1. Claim Ledger — one table per competitor, plus one for us

| # | Claim (verbatim) | Grade | Why | Source + date | Rep action |
|---|---|---|---|---|---|

Rep action is one of: *Concede* (true, don't fight it), *Question* (probe it — C claims), *Counter* (we hold A/B proof against it), *Expose* (D — lay the landmine with a question), *Stop saying* (our own C/D claims).

### 2. SWOT — both companies
Strengths, Weaknesses, Opportunities, Threats for us and for each competitor. 3–5 bullets per cell, each bullet tagged with the grade of the evidence behind it. Our own weaknesses stay in; a flattering SWOT is useless to a rep who will hear the objection anyway.

### 3. How [Your Company] Wins
5 strategies / unique selling points. For each: the claim [grade], the proof the rep can send, and the sentence the rep can say out loud. Only A/B claims qualify. If fewer than 5 qualify, say so and list what proof would unlock the rest — that gap is itself a finding.

### 4. Topics to Avoid
3 areas where the competitor holds A-grade strength and we should not invite comparison. For each: why they win there [grade + source], and the redirect — the topic to steer toward instead.

### 5. Objections & Counters
5 objections a prospect will raise in the competitor's favor. Acknowledge → Reframe → Proof [grade] → Question. Each ends with a question that moves the deal, not a mic-drop.

### 6. Feature Comparison Chart

| Feature | [Competitor] | [Your Company] | Notes / Evidence |
|---|---|---|---|

6–8 rows on the criteria the ICP actually buys on. Cells = ✅ / ⚠️ / ❌ + grade letter. Every ❌ on a competitor carries a source or a `?`.

### 7. Landmines to Lay
4 true limitations per competitor, drawn from D-grade claims and ❌ cells, each written as a **discovery question the rep asks the prospect** ("How are you planning to handle X once you scale past Y?"). The question does the work; the rep never disparages.

### 8. Follow-Up Plays
5 actions (case study, tailored demo, proof doc, reference call, executive alignment), each with when to use it and ONE example email — subject line + ≤100 words, plain human tone.

### 9. Head-to-Head: Why Choose [Your Company]

| Comparison Aspect | [Your Company] | [Competitor] | Reason to Choose [Your Company] |
|---|---|---|---|

6 rows on buying criteria (total cost, time-to-value, support, security, roadmap velocity, ecosystem) — not features already in section 6. Reasons cite grades.

### 10. Marketing Plays
What marketing does with the same evidence. Four short blocks:
- **Positioning gaps they left open** — from the competitor's C and D claims: the category claim nobody has proven, the promise their reviews contradict. Name the angle we can own and the proof we'd need to own it.
- **Fix our own house** — from our C/D claims: what to stop saying on the website today, and the one proof asset (case study with methodology, published customer count, certification) that moves our headline claim to A.
- **Content to build** — a comparison page grounded only in A/B claims; a "questions to ask any vendor" guide built from the landmines; a customer proof piece for the weakest-evidence win reason.
- **Trigger campaign** — if the competitor made a move in the last 90 days (pricing model change, acquisition, outage, credits/metering), the one campaign angle and the audience it targets.

### 11. Talk Track Reps Can Trust
The piece reps actually carry into the call. **Only A and B claims are stated as fact. C claims appear only as questions. D claims appear only as discovery questions that let the prospect find the problem themselves.** Every line carries its grade in brackets.

- **Opener / positioning** (2 sentences) [grade]
- **3 discovery questions** — each surfaces a competitor C/D claim without naming the competitor
- **3 proof points** — A/B only, each with the proof the rep sends afterward
- **Landmines** — one question per competitor D claim (from section 7)
- **Objection responses** — from section 5, condensed to what fits on a call
- **Do-not-say list** — our own C/D claims, plus the section 4 topics
- **Close** — the one ask that advances the deal

### 12. Recent Moves & Watch List
Competitor moves in the last 90 days, dated. What to monitor. Re-run on any pricing change, launch, or funding event, and in 90 days regardless.

### 13. Sources & Open Questions
Numbered sources with dates. Then every `?` and `[USER UNKNOWN]` item as a punch list — what PMM confirms before this goes to the field.

## PHASE 4: Close the loop
Offer: (a) run the remaining competitors, (b) a one-page rep card from sections 0 and 11, (c) rerun after the user supplies proof for any of their own C claims. If a live deal was named, offer to tailor the talk track to that prospect's stated criteria.

---

## Edge cases
- **Only a competitor name, no URL:** search, confirm the right company by showing its URL, proceed.
- **Giant platform competitor (Salesforce, Microsoft, HubSpot):** grade only the competing product line, not the company.
- **Stealth or tiny competitor with no public footprint:** say so, grade what exists, lean on the user's win/loss input, and label the pack Low confidence.
- **User asks you to grade a competitor harder than the evidence supports:** decline that grade, explain that one wrong claim on a call costs the rep the deal, and offer the grounded version.
- **No web access:** say so up front, grade from user-supplied material only, and mark every grade `[UNRESEARCHED]`.

## Worked example (abbreviated)

**Intake:** User: "We're acme-crm.com, RevOps CRM for 50–500 person SaaS. We lose to RivalCRM on price, win on reporting. Their line that hurts is 'the only CRM with native AI forecasting.'"

**Grading:** RivalCRM "only CRM with native AI forecasting" → **D**: three competitors' docs show native AI forecasting [sources]; G2 reviews (n=27) call Rival's forecasting "a black box." Acme "trusted by 200+ RevOps teams" → **B**: logo wall shows 40; "200+" is stated but not listed — tell user what would move it to A (a published customer count or case-study index).

**Talk track excerpt:**
- Discovery: *"When your CFO challenges a forecast number, how does your team explain how the model got there?"* [targets Rival D]
- Proof point: *"Our forecast is auditable to source — here's the Northwind case study, 40% fewer forecast revisions."* [A]
- Do-not-say: "200+ teams" until proof is published [B → hold]

What does the Competitor Intelligence Agent skill do?

Two-mode competitive intelligence agent. Quick Strike grades a competitor's top claims from two URLs in about three minutes and returns a one-screen card. Full Pack runs a guided intake interview, harvests 8 to 12 claims per company, grades each on an A to D evidence scale, and delivers a complete pack: evidence scorecard, claim ledger, SWOT, objections, landmines, follow-up emails, marketing plays, and a talk track where only A and B claims are stated as fact.

Where does the Competitor Intelligence Agent skill come from?

Submitted by a reader of The Revenue AI Report. Reviewed and published as received.. It was written by The Revenue AI Report against a 12 criterion quality rubric and graded in an independent scoring pass.

Why was the Competitor Intelligence Agent skill chosen for this library?

It fixes the two ways reps lose competitive deals: repeating competitor marketing as fact and making claims they cannot back. Every output line carries an evidence grade, so a rep can defend it in front of a skeptical CFO. The Quick Strike mode matches how deals actually move: a call is today, not next quarter.

When should the Competitor Intelligence Agent skill not be used?

Do not use it for: write a general marketing plan Or: score our own brand assets

How do I install the Competitor Intelligence Agent SKILL.md file?

Download the file, create a folder named exactly competitor-intelligence-agent inside your agent's skills directory, and save the file inside it as SKILL.md. The agent loads it when a request matches the description.