COMPLIANT PROMOTION ONLY. Never help with vote manipulation in any form: vote rings/pods, upvote-exchange sites, bought or incentivized votes, fake/sock/clustered accounts, asking people to click an upvote link, or "please upvote us" solicitation. Paying anyone to hunt or send traffic is also prohibited and risks removal/permanent ban. If asked for any of these, decline that part, explain the risk, and redirect to compliant tactics. Every message must pass the banned-phrase scan in references/policy-compliance.md.
GROUND EVERY WORD IN THE REAL PRODUCT. Read the product before writing one tagline — .ulpi/launch/ positioning.md if present, else the README, the live landing page (via browse), .ulpi/design/, and a .claude/.cursor project-context.md. No generic placeholder copy. If you can't tell what the product does and who it's for, ask.
COMPOSE THE SHARED SKILLS, DEGRADE GRACEFULLY. At the delegated steps, invoke launch-copy, launch-outreach (mode product-hunt), and launch-analytics (source producthunt). If a companion isn't installed, tell the user the one-line install (npx skills add https://github.com/ulpi-io/skills --skill <name>) and fall back to the built-in reference so the launch still completes — never hard-fail.
HONOR EXACT, CURRENT FIELD SPECS. Limits and dimensions change — treat references/field-specs.md as authoritative; when a number is critical and might be stale, verify against the live PH submission flow via browse.
RUN THE PRE-FLIGHT GATE before declaring the package done (references/preflight-gate.md). Several checks are mechanical counts. If one box can't be honestly ticked, the launch isn't ready — fix and re-run.
SET REALISTIC EXPECTATIONS. PH is a credibility/visibility amplifier, not primarily customer acquisition, and most submissions are never featured. Never promise virality, a rank, or Product of the Day.
WRITE THE PACKAGE TO DISK under .ulpi/launch/product-hunt/ (and the shared brief to .ulpi/launch/positioning.md). Produce durable, paste-ready artifacts — not advice that scrolls away.

launch-product-hunt

Inputs

$request: the product to launch, a launch date/stage, or a specific PH goal (e.g. "write my tagline and first comment", "plan launch day"), plus optional links.

Goal

Produce a complete, compliant, paste-ready Product Hunt launch package, grounded in the real product, under .ulpi/launch/product-hunt/ (+ the shared .ulpi/launch/positioning.md):

LISTING.md — every PH field, filled and paste-ready (name, tagline options with one chosen, description, topics, gallery shot-list, demo-video outline, the maker's first comment, UTM-tagged links).
PLAN.md — the T-minus countdown, the hour-by-hour Pacific-time launch-day runbook, the hunter/date decision, and the post-launch plan.
OUTREACH.md — supporter segments and paste-ready, PH-compliant messages + social posts.
CHECKLIST.md — the pre-flight gate result and the launch-day checklist.

Step 0: Discovery & companions

Resolve, asking only what you can't determine: the product (name, what it does, who for, live URL); stage & date (imminent or weeks out; date chosen?); audience/reach; team; hunter (self-hunt default). Note which companion skills are installed (launch-copy, launch-outreach, launch-analytics); if any are missing, mention the install command but proceed either way (fallbacks exist). Ask at most a couple of focused questions. Success criteria: product, stage/date, and reach are known.

Step 1: Ground the product → `.ulpi/launch/positioning.md`

Reuse .ulpi/launch/positioning.md if it exists. Otherwise build it from .ulpi/design/DESIGN.md, a .claude/.cursor project-context.md (Sections 1–9), the README/docs, and the live landing page (visit it with the browse skill — extract the real value prop, features, screenshots, proof points). If browse or a site is missing, ask the user for the key points. Distill: one-line what-it-is, ICP, core value in the user's words, top 3 differentiators, proof points, category, links — nothing invented. Save it as the shared source of truth. Success criteria: a brief no asset will contradict.

Step 2: Mechanics, timing & realistic targets

Load references/ph-mechanics.md. Decide and record (into PLAN.md): the date/day (Tue–Thu default, 12:01am PT; weekend only as a deliberate low-competition play), the hunter decision (self-hunt by default; never paid), and an honest outcome range (featuring/Product of the Day not guaranteed). Success criteria: a concrete date, a hunter decision, and a realistic target — no hype.

Step 3: Write the listing → `LISTING.md`

Hand the PH asset profile (references/copywriting.md) and the exact limits (references/field-specs.md) to the launch-copy skill to draft the tagline options (≤60, one recommended), description (≤500), gallery slide headlines, and the maker first comment — all bound to positioning.md, in PH voice, with zero upvote asks. Fallback: if launch-copy isn't installed, draft them directly from references/copywriting.md. Then PH-owned: assemble LISTING.md, add the gallery shot-list (image 1 = the scroll-stopping poster + caption, then 3–5 value slides — 4–6 total at 1270×760), the demo-video outline, topics (≤3), makers, any offer (tied to trying, never voting), and the links. Success criteria: every field filled, on-brand, within spec (characters/dimensions counted), paste-ready — nothing says "[insert tagline]".

Step 4: Plan the ramp & launch day → `PLAN.md`

Load references/prelaunch-playbook.md and references/launchday-runbook.md. Write the T-minus countdown (scaled to the real timeline — build the off-platform supporter list since PH's native teaser pages are gone, warm the audience, finalize assets, schedule the launch) and the hour-by-hour launch-day runbook (Pacific): 12:01am publish + first comment → notify in waves → reply to every comment fast → midday push → final-hours push → day-wrap thanks. Assign team roles if there's a team. Success criteria: a teammate could run the day from PLAN.md without improvising.

Step 5: Outreach & amplification → `OUTREACH.md`

Invoke the launch-outreach skill with compliance mode product-hunt, positioning.md, and the launch date — it produces the segmented plan and paste-ready, compliant messages (teaser DM, waitlist email, launch-day email waves, community heads-up, 1:1 DM, X thread, LinkedIn) plus the wave schedule. Fallback: if it isn't installed, draft them from references/policy-compliance.md (the safe-vs-unsafe phrasing and templates there). Success criteria: every message is paste-ready, personalized, and passes the no-vote-solicitation scan.

Step 6: Wire measurement → tag links & events

Invoke the launch-analytics skill for this launch — channel product-hunt (the on-disk slug, so it writes .ulpi/launch/product-hunt/analytics.md alongside the rest of the package) with utm_source producthunt (the GA value) — to set the UTM scheme, tag every launch link in LISTING.md and OUTREACH.md, and wire signup/activation tracking. Fallback: if it isn't installed, apply a minimal inline UTM scheme (utm_source=producthunt&utm_medium=<placement>&utm_campaign=<launch>) and write the UTM map + signup/activation plan to .ulpi/launch/product-hunt/analytics.md so the artifact exists either way. Success criteria: every link is attributable and signups are trackable — not just upvotes.

Step 7: Pre-flight gate → `CHECKLIST.md`

Run references/preflight-gate.md end to end and write the result into CHECKLIST.md. Several items are mechanical: tagline ≤60 (counted), description ≤500, gallery count/dimensions, first comment present, zero banned vote-soliciting phrases across all outreach, links UTM-tagged, topics ≤3, supporter list realistic. Then append a short launch-day checklist — the launchday-runbook.md cadence distilled to tickable go-live steps (publish 12:01am PT, post first comment, notify wave 1, reply to comments, midday push, final-hours push, day-wrap thanks) — so CHECKLIST.md carries both the readiness gate and the day-of checklist. Fix any failing box and re-run. Success criteria: every box ticked; the package is internally consistent and compliant.

Step 8: Post-launch plan → append to `PLAN.md`

Load references/postlaunch.md. Append: claim/embed the badge if earned, ride the leaderboards/ newsletters/Orbit reviews, convert the traffic spike (emails, onboarding, retargeting via launch-analytics), send thank-yous, repurpose into social proof, the re-launch rules (6+ months + a significant update), and the "if it's flopping" contingency (keep engaging, never panic-blast or beg). Success criteria: the user knows exactly what to do the moment the 24-hour window ends.

Guardrails

Coach compliant promotion only; never manipulate votes; decline and redirect if asked. No paid hunts.
Ground all copy in the real product; never ship invented features, fake metrics, or generic taglines.
One positioning, bound across every field and message; reuse .ulpi/launch/positioning.md.
Compose the shared skills; if one is absent, install-hint + fall back — never hard-fail.
Honor exact, current field specs; verify a critical number via browse when it might be stale.
Set realistic expectations; never promise a rank, featuring, or virality.
This skill specifies the gallery and video; it does not render images or edit video.
Write the package to disk; don't leave the launch plan as ephemeral chat.

When To Load References

references/ph-mechanics.md — ranking/featuring, the day window, awards, best-day timing, realistic outcomes (Step 2).
references/field-specs.md — exact current specs for every submission field (Step 3). Authoritative.
references/copywriting.md — the PH listing asset profile passed to launch-copy, and the inline copy fallback (Step 3).
references/prelaunch-playbook.md — the T-minus countdown, supporter-list building, self-hunt vs hunter, scheduling (Step 4).
references/launchday-runbook.md — the hour-by-hour Pacific runbook, waves, engagement, roles, hard rules (Step 4).
references/policy-compliance.md — prohibited behavior, shadowban triggers, the banned-phrase scan, and the outreach fallback (Steps 5, 7).
references/preflight-gate.md — the blocking readiness + compliance gate (Step 7).
references/postlaunch.md — badges, leaderboards, Orbit reviews, converting the spike, thank-yous, re-launch rules, and the flop contingency (Step 8).

Output Contract

Write under .ulpi/launch/ and report:

.ulpi/launch/positioning.md — the shared grounded brief (created or reused)
product-hunt/LISTING.md — name, tagline options (one recommended), description, topics, gallery shot-list, demo-video outline, maker first comment, UTM-tagged links
product-hunt/PLAN.md — the countdown, the hour-by-hour runbook, the hunter/date decision, the post-launch plan
product-hunt/OUTREACH.md — supporter segments and paste-ready compliant messages + social posts
product-hunt/CHECKLIST.md — the pre-flight gate result (every box ticked) and the launch-day checklist
product-hunt/analytics.md — the UTM map + signup/activation tracking (from launch-analytics)
a realistic expectation set and the compliance affirmation (zero vote-soliciting language in any message), plus which companion skills were used vs fell back to built-ins