Rymac Paid Traffic Landing Page

Paid-Traffic Landing Page

Build a landing page whose only job is converting one ad click into one action. Headline quality is ~80% of the outcome; simplicity is the other 20%. Everything else is deleted, not designed.

Reference files (read at the step that needs them):

• ${CLAUDE_SKILL_DIR}/references/niche-research.md — 12 intake questions, the 20-minute pain-mining sprint, sophistication stages, pain→copy mapping
• ${CLAUDE_SKILL_DIR}/references/headlines.md — awareness levels, 13 rules, 16 formulas, subheadline formulas, message match
• ${CLAUDE_SKILL_DIR}/references/page-structure.md — structure rules with conversion data, single-screen anatomy, CTA/trust rankings, mobile rules
• ${CLAUDE_SKILL_DIR}/assets/page-template.html — the single-screen HTML shell with {{TOKENS}} to fill (read it, fill it, save as the deliverable; don't rebuild the layout from scratch)

Workflow

Copy this checklist and tick it off:

• [ ] 1. Intake — parse $ARGUMENTS; collect what's missing
• [ ] 2. Mine the niche (research sprint, parallel subagents)
• [ ] 3. Diagnose awareness + sophistication
• [ ] 4. Write the copy elements, headline first
• [ ] 5. Build the page from the template
• [ ] 6. Run the conversion gate — fix anything that fails
• [ ] 7. Deliver page + copy doc with A/B alternates

1. Intake

Needed before writing: niche/avatar, offer, traffic source + the ad's promise (for message match), CTA goal (lead form / booking / click-through), brand name + accent color. Take what $ARGUMENTS and context give; ask the user only for genuinely unknowable items (the ad copy, the brand color). If no ad exists yet, note that the H1 should become the ad headline later — write them as one unit.

2. Mine the niche

Read ${CLAUDE_SKILL_DIR}/references/niche-research.md and run its sprint. Fan out 2–3 parallel research subagents (Reddit/forums · reviews of competitors or current solutions · Meta Ad Library/competitor headlines) and have each return verbatim quotes tagged pain / desire / objection / failed-alternative / trigger. Rank by frequency × emotional charge. Output: 1 dominant pain, 1 dream outcome, top 3 objections, top 2 failed alternatives — in the prospect's exact words.

3. Diagnose

• Awareness of the traffic (cold social ≈ problem/solution-aware; search ≈ solution-aware; retargeting ≈ product/most-aware) — sets which headline formulas apply.
• Sophistication of the market (stage 1–5 from competitor ads) — write one stage ahead of the loudest incumbent; never repeat a worn-out claim.

4. Write the elements (headline first)

Read ${CLAUDE_SKILL_DIR}/references/headlines.md. In order:

1. Headline — write 20 candidates from the 3 best-fitting formulas; if ad copy exists, the verbatim ad promise is always candidate #1. Cut to the strongest 3 by the rules (≤8 words, specific, visualizable, outcome-led, prospect's own words). Pick one; keep the other two as A/B variants.
2. Subheadline — mechanism + proof + kills the #1 objection. 10–30 words.
3. 3 bullets max — ≤7 words each, specific numbers, mined language. Skip entirely if headline+sub carry the page.
4. CTA — [verb] + my + [outcome] ("Get My Free Quote"). Never "Submit"/"Learn More".
5. Risk reversal — one line under the button, aimed at the #1 mined objection ("Free — no credit card" / guarantee / "Takes 30 seconds").
6. Trust line — ONE element: stars+count, micro-testimonial with name + number, or 3–4 known logos. Only what's real — never fabricate reviews, counts, or testimonials; if nothing real exists, use effort reassurance instead.

5. Build the page

Fill ${CLAUDE_SKILL_DIR}/assets/page-template.html (copy it into the project/output directory first). Lead form → Variant B with the fewest fields follow-up can survive on (3 max); booking/click-through → Variant A. Set the accent color; keep it the only saturated hue on the page. No additions: no nav, no sections below the fold, no images unless one visual directly shows the deliverable/outcome, no external scripts except the ad pixel if provided.

6. Conversion gate

Verify every line; fix and re-check anything that fails:

• [ ] H1 restates the ad's promise (message match) — or is flagged to become the ad headline
• [ ] H1 ≤8 words, outcome-led, passes the 5-second test (what / for whom / what to do)
• [ ] Copy reads at 5th–7th grade level
• [ ] Exactly ONE decision: one CTA goal, zero other links (Privacy/Terms ghost-text footer only)
• [ ] Headline + subhead + CTA + risk line all visible in a ~390×844 viewport without scrolling; CTA ≥48px tall, full-width on mobile, never grey
• [ ] Form (if any) ≤3 fields, correct input types
• [ ] One trust element, specific and true, adjacent to the CTA
• [ ] No render-blocking or third-party requests; page is a single self-contained file

7. Deliver

Hand back: the HTML file path, the chosen headline with the two A/B alternates, and a short copy doc (all elements + the mined quotes that justify them) so ads can be written with the same language. Recommend one page per ad angle — offer to spin variants by swapping H1/eyebrow.

Rules

1. The headline is the product. Never spend more effort on layout than on headline candidates.
2. One page, one pain, one decision. Cut a second idea rather than fit it.
3. Prospect's verbatim words beat anything invented — if a phrase wasn't mined or given, question it.
4. Never fabricate proof (reviews, star counts, names, "as seen in").
5. Delete before adding: when the fold is tight on mobile, bullets go before the CTA does.

Source & license

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

• Author: oathdriven
• Source: oathdriven/rymac-skills
• License: MIT

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