Product messaging

Builds a 10-component messaging library from website and product research. Output ships as the source of truth for all downstream marketing assets — landing pages, sales enablement, LinkedIn content, outreach. Knowledge type: messaging (per .claude/rules/ontology.md); maturity: emergent → validated after team review → canonical when locked. Visual phase map → the premium reference.

When to run

Invoke when the user asks for: product messaging for [URL/company], messaging library for [product], extract messaging from [website], product messaging framework, capabilities and benefits for [company], what are [product]'s differentiators?, pain points and capabilities for [URL]. Do NOT invoke for: competitor analysis only (use /competitor-research), landing page copy directly (use /landing-page-copy — run this first), ICP research only (use /icp-behavioural), or single-feature questions (answer directly without full framework).

The Iron Law: no messaging output without source verification. Every claim cites URL + access date or is marked [Not available]. Every quote is verbatim. Every consequence chain traces 1st→2nd→3rd order. Full guardrails + red flags + anti-hallucination rules → the premium reference.

Inputs

Required:

• website URL — primary product website (verify it loads).
• product name — exact product/company name (confirm if ambiguous, e.g., "Bolt" could be ride-share, fintech, or bolt.new).

Recommended (improve quality):

• target ICP context — focuses messaging on relevant segments.
• competitor context — sharpens differentiators (use /competitor-research output if available).
• internal docs — provides claims not on website.
• customer quotes — fills gaps in testimonial coverage.

Upstream skill outputs (if available, read first):

• positioning (primary) — frames Description and core messaging blocks.
• icp-behavioural — enriches pain points and benefits with VoC data.
• competitor-research — sharpens status quo and differentiators.
• tov-guidelines — applies tone to messaging.

If website URL is missing, ask. If product name is ambiguous, confirm before starting.

Steps

1. Validate inputs → verify URL accessible, product name confirmed, ICP context confirmed if not obvious. Pull upstream skill outputs (positioning, icp-behavioural, competitor-research) into context if available.
2. Phase 1 — Discovery research → the premium reference. Fetch core pages (homepage, features, pricing, customers, about — 5+ pages with URLs + access dates). Search external data (G2, testimonials, vs-pages) per Exa protocol (.claude/rules/exa-protocol.md). Extract branded feature names verbatim before structured extraction begins. Detailed search/scraping patterns → the premium reference.
3. Phase 2 — Structured extraction → the premium reference. Extract all 10 components in order: Description → Status quo & alternatives → Pain points (with consequence chains) → Capabilities → Functional benefits → Emotional & social benefits → Features (branded names) → Cost of inaction → Common objections → Core messaging blocks. Frameworks + descriptor counts + link graph → the premium reference.
4. Phase 3 — Verification & gaps → the premium reference. Source-verify every claim (URL + access date), confirm verbatim quotes, assign confidence levels (High/Medium/Low → [VERIFIED]/[INFERRED]/[ESTIMATED]), document data gaps + recommendations as Component 10.
5. Apply attribution standards → per .claude/rules/ontology.md: [VERIFIED: source_type, reference], [INFERRED: from X + Y], [ESTIMATED: reasoning], [UNAVAILABLE]. Quality threshold for client-deliverable strategy outputs: ≥60% verified, ≤10% estimated.
6. Self-evaluate against quality gates → the premium reference. Run completeness, evidence-quality, and guardrail checks. Answer self-roast questions honestly.
7. Write to client folder per output template → the premium reference. File path: messaging/MMYY-messaging.md (or per client CLAUDE.md folder map). Header includes skill name, generated date, font (Inter), version.
8. Push to Notion (Product Messaging Database) and Google Docs (client_folder/strategy/) per push targets in frontmatter. For refresh runs, UPDATE existing pages rather than duplicating.
9. Offer iteration prompts post-delivery → the premium reference. If user signals approval ("great messaging" / quick approval), offer to save as a reference example under the premium reference.

What good looks like

Evaluations (binary pass/fail before declaring "done")

• All 10 components present in correct order (or explicit [Not available] per component with reason).
• Status quo includes Manual/DIY + at least one named competitive alternative.
• Every pain point has a complete 1st→2nd→3rd order consequence chain (not cut short).
• Every pain point links forward to a capability; every capability links to a branded feature name.
• Every functional benefit includes a verbatim customer quote or explicit "Not available" with reason.
• Emotional + social benefits section complete: 2 emotional + 2 social.
• Cost of inaction section quantified (daily/weekly/monthly cost stated, not abstract).
• Common objections section complete (3-5 objections with root cause + Acknowledge/Reframe/Evidence response).
• Core messaging blocks complete: tagline ≤7 words, elevator pitch (1 sentence), 3-bullet value prop, proof point, audience-segmented messages.
• Every claim has source URL + access date; every quote verbatim; confidence level assigned (High/Medium/Low → [VERIFIED]/[INFERRED]/[ESTIMATED]).
• ≥60% [VERIFIED] confidence; ≤10% [ESTIMATED] (per ontology threshold for client deliverables).
• Data gaps section non-empty if any component is incomplete or low-confidence; recommendations provided for filling each gap.
• Source appendix lists ALL referenced URLs with access dates.
• Output title is # Product messaging library: [Product Name] exactly — no aliases.

Pre-slim original

Pre-slim SKILL.md (1,048 lines, v2.1) archived at .claude/skills/_archive/messaging/SKILL-pre-slim-20260429.md. See the premium reference ("Changelog") for the v2.2 entry documenting the slim.

---

Final ship gate

Run /premortem --output before ship. See /premortem skill for the 5 execution domains (will-it-resonate / will-it-convert / will-it-stay-on-brand / will-stakeholder-push-back / will-it-degrade-over-time) and output template.

Trivial-case escape: ## Premortem\nNo failure modes — trivial change satisfies the contract for genuinely trivial outputs.