JPeetz/agent-skills / skills/business-strategy/go-to-market-launch
go-to-market-launch
VerifiedUse this skill when planning a launch, beta, or GTM push.
Go-to-Market Launch
A launch is a decision about scope and channel, not a press release. This skill produces a launch plan a team can execute on a calendar: a go/no-go verdict, a tier, a channel map that ends in owned assets, a narrative, a runbook with rollback criteria, and an activation target that survives week two.
Output is a plan document, not copy. Ad creative and video assets are produced by other skills (see boundaries).
When to Use
- Shipping a new product, a paid tier, or a feature that changes what the product is
- Choosing how loud to be: quiet ship vs private beta vs full public push
- Planning Product Hunt, Hacker News, or a directory/registry listing
- A launch date is fixed and you need a day-of runbook with abort conditions
- Post-launch: signups arrived and activation did not, and you need to find out why
- Converting a shipped repo or internal tool into a public release with an audience
Do Not Use For
| Need | Use instead |
|---|---|
| Ad copy, hooks, scripts, A/B test plans for X/TikTok/IG | astra-campaign |
| Rendering a launch video (TTS, talking head, captions) | agent-n-video-pipeline |
| Setting the price itself | pricing-strategist |
| Ongoing content calendar after launch week | content strategy skill |
| Deciding whether the idea is worth building | app-discovery-scrutiny |
| Deal desk, discounting, sales compensation | revenue ops |
If the ask is "write my launch tweets," that is creative production. This skill decides who to tell, in what order, with what proof, and when to stop.
Step 1 — Readiness Gate
Run this before any channel work. A loud launch of an unready product burns the one attention window you get with a channel; the second attempt is always weaker than the first.
The SLC test (Simple, Lovable, Complete)
Launchable ≠ feature-complete. It means one job, done as a whole experience.
| Criterion | Passes when | Fails when |
|---|---|---|
| Simple | It does one clearly stated thing | The pitch needs "and" three times |
| Lovable | A target user would choose it over their workaround | Users tolerate it to "give feedback" |
| Complete | No visible stub inside that one scope | "Coming soon" sits in the main flow |
Two opposite failure modes both kill launches:
- Stealth mode — polishing in private, waiting for perfect. Procrastination in a nice suit. Nothing ships, nobody learns you exist.
- Just one more feature — every date slips for one more thing. The launch becomes a permanent internal project.
Middle path: cut scope until one thing is lovable and complete, ship that, then launch the rest as updates. A launch you can repeat beats a launch you never do.
Launch blockers (hard stops)
Any one of these true → do not run Tier 2:
- Core flow breaks for a new user on a clean account (no demo data, no hand-holding)
- Signup, payment, or email delivery has no one on call
- No way to measure activation (see Step 7) — you would be launching blind
- Support route is undefined (no inbox, no docs, no owner)
- Terms, privacy policy, or data-handling claim is missing where required
- Rollback is undefined: you cannot name how to turn the launch off
Readiness verdict
Write one line: SHIP / SHIP SMALL / HOLD, plus the single biggest risk. If you cannot name the biggest risk, you have not looked.
Step 2 — Launch Tiering
Pick the tier from readiness × consequence of being wrong, never from how excited the team is.
| Tier 0 — Quiet ship | Tier 1 — Beta | Tier 2 — Full push | |
|---|---|---|---|
| Trigger | Feature inside an existing product; reversible; low blast radius | New surface, real users, still-shifting scope | New product/category, or a moment competitors cannot copy quickly |
| Audience | Existing users, changelog readers | Named waitlist, 20–100 invited accounts | Everyone, all channels, one date |
| Channels | Changelog + in-app note + one owned post | Email to list, 1–2 rented posts, docs | ORB full set + PH/HN + borrowed placements |
| Assets | Screenshot, 5-line note | Landing page, demo GIF, onboarding email | Landing page, demo, 60–90s video, press kit, comparison page |
| Success metric | Adoption among exposed users | Activated invited cohorts | Activation rate + qualified pipeline |
| Cost of failure | Low — retry next week | Medium — invitees remember | High — channel credibility is spent once |
| Reversibility | Instant | Feature-flag off | Public; only mitigable |
Rules:
- Default down. Tier 2 is earned by a passing gate, not requested by enthusiasm.
- One Tier 2 per quarter, per channel. Channels remember. Two PH launches inside eight weeks splits your own vote.
- A Tier 0 that goes well is the argument for Tier 1 — use the adoption data as the proof point.
- Never tier up mid-launch. Escalating on day one because signups spiked is how rollback criteria get ignored.
Step 3 — ORB Channel Framework
Three channel classes, distinguished by who controls distribution.
Owned — you control access
Email list, blog, docs site, changelog, branded community (Discord/Slack), the product itself.
- Compounds; no algorithm, no pay-to-play, no platform risk.
- Start with one or two matched to the audience: writers → blog; operators → email; developers → docs + community.
- An owned channel is only owned if you can export the list.
Rented — visibility without control
X, LinkedIn, YouTube, Reddit, app stores, marketplaces, directories.
- Algorithms change, reach decays, rules are enforced unevenly.
- Use 1–2 where the audience already is; each post must have a path back to owned.
- Rented reach is a spike, not a base. A launch that lives only here ends on day three.
Borrowed — someone else's audience
Guest posts, podcasts, newsletter features, co-webinars, creator reviews, speaking slots.
- Fastest credibility shortcut; also the least controllable timing.
- Ask for a specific slot tied to your date, not "sometime soon."
- A borrowed placement without an owned capture step is a vanity metric.
The rule: everything routes to owned
Every rented post, borrowed mention, PH listing, and directory entry must terminate in an owned asset — email capture, waitlist, docs signup, or account creation.
Rented/Borrowed attention → landing page (single CTA) → owned list/account → onboarding → activation
If a channel cannot route anywhere owned, treat it as brand spend, not launch spend, and say so in the plan.
Channel map (fill before launch)
| Channel | Class | Audience fit | Owned destination | Owner | Date |
|---|---|---|---|---|---|
| Changelog post | Owned | existing users | in-app | T-1 | |
| Email to list | Owned | warm | product | T-0 | |
| X thread | Rented | landing page | T-0 |
A map with twelve rented entries and no owned destination is a red flag.
Step 4 — Launch Narrative
The narrative is four sentences and one demo. If it needs a deck to be understood, it is not ready.
Positioning statement
For [specific user] who [situation/pain], [product] is a [category] that [single outcome]. Unlike [alternative], it [one defensible difference].
Keep the category recognizable. Inventing a category on launch day forces the audience to do work they will not do.
Proof points
Three, ranked. Each must be checkable by a skeptic in under two minutes:
- Demo-able — they can see it work (best)
- Measured — a number you can source and defend (
Nruns,X% faster,$Ysaved) - Testimonial — a named or clearly-described user with a specific before/after
Rules: no unlabeled projections, no "up to," no borrowed logos you have not earned. One overstated claim converts a technical audience into a hostile one.
Objection table
Write the six objections you would raise if you were the buyer. This is the highest-leverage page in the plan — it becomes the FAQ, the PH comment replies, and the HN answers.
| Objection | What it really means | Response (proof point + link) |
|---|---|---|
| "This is a toy / demo" | needs to see durability | |
| "I can build this in a weekend" | needs the cost of the hard part | |
| "Why not ?" | needs the specific trade-off | |
| "Is my data safe?" | needs the data path, plainly | |
| "What does it cost me later?" | needs the pricing shape, not a number | |
| "Who else uses it?" | needs one credible peer |
Demo
- 60–90 seconds, no voiceover intro, outcome visible in the first 5 seconds
- Recorded, not live, for the launch page; shows the job, not an interface tour
- Works on a fresh browser profile with no cached session — test this literally
Step 5 — Channel Playbooks
Product Hunt
Prep beats luck. The listing is a landing page with a vote counter.
T-14 to T-7
- Confirm the hunter or self-hunt; pick a launch date with no obvious giant shipping
- Write the tagline as the positioning statement's outcome clause, ≤60 chars
- Assets: gallery of 4–6 frames (outcome first, settings last), 30–60s demo, thumbnail legible at 100px
- Seed 15–30 people who will genuinely try it; ask for feedback, never for votes
- Verify the first-comment draft: what it is, who it's for, what's different, one link
T-0
- Launch 12:01am PT; the day runs on US waking hours
- Reply to every comment within 30 minutes, all day; answers, not thanks
- Do not link-drop in unrelated communities — that is the fastest way to get delisted
- Route every visitor to a page with email capture, not straight to signup
T+1 to T+7
- Thank-you note with the outcome; convert the listing traffic into the owned list
- Publish the "what we learned" post — it outperforms the launch post most of the time
Realistic expectations: a #1 day is a spike of low-intent traffic. Judge it on activated signups, not upvotes.
Hacker News
The only channel where the comment section is the launch. Technical credibility or nothing.
- Show HN only for something people can try immediately; title = "Show HN: ()", no adjectives
- Post 6–9am ET on a weekday; be present for the first two hours
- Expect the harshest objection first ("this already exists", "what's the catch"); answer with mechanism, not marketing
- Never ask for upvotes, never use alt accounts, never argue with a hostile top comment — one calm technical reply and move on
- If it does not take, do not repost the same thing within weeks; ship something and try again later
Directories and registries
Low effort, long tail, compounding SEO — roughly an afternoon total.
- Pick directories your audience actually browses (AI tool lists, dev registries, niche industry indexes), not the 200-site spray lists
- Reuse one asset set: 60-char tagline, 160-char and 300-word descriptions, logo, 3 screenshots, pricing shape
- Every listing links to an owned landing page with a UTM; revisit quarterly, since stale entries decay but still send traffic
Step 6 — Launch-Day Runbook
One page, times in one timezone, named owners. Unnamed tasks do not happen.
| Time | Action | Owner | Done when |
|---|---|---|---|
| T-7 | Assets frozen; landing page and onboarding QA'd on a clean profile | no P0 bugs open | |
| T-3 | Email drafted and test-sent to 3 addresses incl. one external | inbox + mobile render OK | |
| T-1 | Staging → prod deploy; smoke test signup, payment, email, docs links | all four pass | |
| T-1 | Status page / support inbox monitored; on-call named for 48h | rota posted | |
| T-0 08:00 | Publish post + changelog; flip any flags | URLs live | |
| T-0 09:00 | Email send | delivery rate > 97% | |
| T-0 09:30 | Rented posts; PH listing live if used | links resolve | |
| T-0 hourly | Metrics check: signups, activation, errors, support queue | dashboard open | |
| T-0 18:00 | Day-one retro note: what broke, what converted | written, not verbal | |
| T+1 | Onboarding email sequence live; reply to every inbound | queue empty | |
| T+7 | Activation review against target; decide iterate vs expand | decision recorded |
Rollback criteria
Decide these before launch, in writing. A rollback call made under pressure defaults to doing nothing.
Abort the public push (delist / pause posts / pull the banner) when any of:
- Error rate on the core flow exceeds 5% for 15 minutes
- Signup or payment fails for a whole provider (not a single user)
- A data-exposure or auth issue appears — this is immediate, no debate
- Support queue exceeds 2× the on-call's capacity to answer within 4 hours
- The demo makes a claim the product cannot do; a wrong claim in public is worse than a missing launch
Roll back quietly, communicate loudly. Fix first, then a short factual note: what happened, who was affected, what changed. Silence after a public failure costs more than the failure.
Step 7 — Post-Launch Measurement
Signups measure the channel. Activation measures the product. Only one of them predicts revenue.
Define activation before launch
A user is activated when they [single observable action] within [window].
Good activation events are actions the user takes for their own benefit, and that a non-activated user cannot fake:
- Good: "created and shared one report within 7 days", "ran 3 analyses in first session"
- Bad: "completed onboarding", "invited a teammate" (proxy for intent, not value), "logged in twice"
The report that matters
| Cohort | Signups | Activated | Activation % | Day-7 return % | Source |
|---|---|---|---|---|---|
| PH day | producthunt | ||||
| HN | hn | ||||
| Email list | owned-email | ||||
| Directory | listing |
Read it by channel, not in aggregate. A 40% aggregate that hides 3% from PH and 55% from email tells you where to spend next time.
Cadence
- T+1: funnel drop-off — where do new accounts die?
- T+7: activation by channel; kill the channel that produced noise, not the one that produced fewer, better users
- T+30: retention of the launch cohort vs pre-launch baseline. Did the launch attract a different user than the product is built for?
Signals worth more than signups
- Activated accounts that returned without a reminder
- Inbound support questions that are about use, not access
- Pre-launch list converting at a materially higher rate than cold traffic — proof the narrative lands
Anti-Patterns
- Launching everything at once. One push, then silence for a quarter. Stagger: launch the product, then ship the second act while attention lasts.
- Big launch, no owned capture. Traffic lands on a signup wall, bounces, and is gone forever.
- Signups as the headline metric. A spike with 4% activation is a paid lesson in vanity.
- Tier inflation. Calling it a "beta" while running a full public push — you get the scrutiny of a launch without the forgiveness of a beta.
- Launching a stub because the date was announced internally. Cut scope, keep the date.
- Rollback criteria invented mid-incident. Written in advance or it will not be used.
- Borrowed mentions with no ask. "Let us know if you'd like to cover it" converts to nothing; propose a date and an asset.
- Claiming traction you do not have. Numbers get checked. One fabricated metric ends the relationship with the exact audience you wanted.
- Ignoring the comments. On PH and HN the reply is the marketing; a silent launch reads as an abandoned one.
- Post-launch silence. The week after is where the launch either becomes a habit or a memory.
Verification
Before handing the plan over, confirm:
- A named readiness verdict (SHIP / SHIP SMALL / HOLD) with one stated biggest risk
- The tier is justified by the gate, not by ambition, and is internally consistent (assets, channels, metric)
- Every channel in the ORB map has an owned destination; no orphan rented channels
- The narrative contains a positioning statement, exactly three proof points, and ≥5 objections with responses
- The runbook has owners and times, and rollback criteria are numeric and written before T-0
- Activation is defined as one observable action with a window, and the measurement table is per-channel
- Nothing in the plan depends on a metric you cannot measure on T+7
References
references/sources.md— source skills, frameworks, and reading list behind this one
License
MIT — see LICENSE.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |