Launch planning

A launch run from start to finish

One directory batch with the kit, end to end: brief, batch selection, one platform at a time, evidence-based states, resume rules, and a realistic first result.

Updated

A run has a target, recorded outcomes and a clear next action. Start with 10 new confirmed submissions, with at most 2 platforms in parallel when your machine and browser setup support it. Your agent performs the work; the dashboard saves the plan and progress. Begin after onboarding and the Chrome connection test.

1. Check the product and permissions

Read the existing product brief, identity, permissions and submission records. Use verified features, current pricing, genuine screenshots and your approved public wording. A trial stays a trial. Unknown mandatory facts stay unknown until checked. Confirm the destination account and search for existing listings before creating anything.

2. Plan the run

Choose the number of new submissions and the parallel limit in the local run planner. The recommendation is 10 submissions and a ceiling of 2 workers. Hardware affects concurrency, not how many platforms fit your product. A machine with limited resources, high current load or unknown capacity should start with one worker. Even a powerful machine starts with a ceiling of two.

Two workers require separately controlled browser workspaces and distinct platform assignments. If they share browser focus or cannot keep their work isolated, use one. Never let both select the same platform. Read each candidate’s full playbook and check current eligibility and free access before opening its form. Replace unsuitable or paid routes with suitable alternatives; do not lower your standards to reach ten.

3. Start with an explicit goal

Copy the saved run prompt. If your agent supports /goal, use its goal feature with that text. Otherwise paste it as a normal message. /goal is a host feature, not a universal LaunchRepo command. Provider usage limits, missing permissions and human challenges still apply.

A typical goal is:

Complete 10 new, relevant, free directory submissions for this product. Use at most 2 workers, only with sufficient resources and isolated browser control; otherwise work sequentially. Read the saved run and existing submissions first. Save each confirmed outcome and its evidence, then continue. Replace blocked candidates with suitable alternatives. Stop when the target is reached or explain the exact reason no further progress is possible. Preserve the remaining work for resume.

4. Submit, save, close, continue

For each assigned platform:

  1. Recheck eligibility, free access, required fields, consent and the signed-in identity.
  2. Prepare approved copy and genuine assets. Read fields back and verify that drafts persisted.
  3. Inspect the final total. Skip required payments. If an official badge is required, follow the recorded website permission and verify the public deployment before attesting.
  4. Submit within the owner’s permission and inspect the actual confirmation. A timeout requires checking the account and listing before any retry.
  5. Record the observed state, non-sensitive evidence and next action immediately.
  6. After evidence is saved, close only the window assigned to this submission and its owned verification tabs. Keep personal windows, other workers, the dashboard, unsaved forms and open human handoffs intact. Keep the browser profile for future sign-ins.
  7. Continue until the saved goal is reached or no eligible work can proceed.

5. Count confirmed submissions accurately

Outcome Counts toward the run?
New submitted_pending_review, queued, scheduled or live Yes, once per platform, with confirmation
Later transition from pending to live No additional count
already_listed or an older submission No
Draft, CAPTCHA, 2FA or missing mandatory fact No
Paid route, unsuitable or unavailable platform No

Pending review is a successful submission, not a promise of publication. A scheduled launch must respect the product’s earliest launch date. Partner syndication counts separately only after that partner confirms its own submission. See recording outcomes for the tracker’s evidence requirements.

6. Resume the remaining work

When a session stops, preserve the run and its remaining target, update SESSION.md with the next action and keep human handoffs visible. Start the next session with the saved resume prompt. Reuse standing permissions, recheck unresolved attempts and follow platform review windows without inventing an SLA. Never submit again merely to move up a queue.

If the dashboard shows 6 of 10 submitted, there are four new submissions left. It may also show two platforms in progress and one waiting on you; those are not added to the six. If only six suitable free routes exist, finish with that honest result and the reason the target could not be reached. Ten is a target, not a guaranteed inventory of suitable platforms.

The kit these guides describe

LaunchRepo is a private repository your own coding agent runs: 299 directory playbooks, the workspace scripts, the tracker and the local dashboard. Pay once, launch every product you build.

See pricing