agentsclimarketplace

Run launch

Skill hungv47/meta-skills/skills/meta/run-launch

Run one channel's launch end to end as ONE legible bundle — for Product Hunt, Reddit, or any launch-channel pack. Resolves the channel pack, sequences the existing leaf skills, narrates each step's applied play, stops at the human gate, never auto-publishes. Use when you want the whole launch orchestrated, not just one piece. Not the copy bundle alone (write-launch), the run-of-show alone (plan-campaign), or a generic produce-and-measure loop (run-pipeline).From its SKILL.md

Install
npx -y skills add hungv47/meta-skills --skill run-launch

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 13 stars13 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

6.7 KB, ~1.5k tokens by cl100k_base, as published. Nobody here has run it

Run Launch — Per-Channel Launch Orchestrator

Meta process skill. Threads the per-channel launch chain into ONE legible, gated launch — resolves the pack, sequences the leaf skills, narrates each applied play, stops at the human gate. Orchestrate, don't fuse; never auto-publish (D-8).

Core question: "Did this ship as ONE coherent, channel-native bundle — each step's applied play visible, the pack named — stopping at the human gate, never auto-published?"

The 7 steps, per-step pack binding, the unwired-leaf degrade, the bundle contract: launch-chain-spec.md [SPEC]. Pack contract: CONTRACT.md.

Critical Gates

  1. Orchestrate, don't fuse. Each step dispatches its leaf (research-icp, plan-campaign, brief-graphic/brief-shortform, write-launch/write-social, write-outreach, publish-social, measure-results); this skill owns only sequencing, pack narration, and the bundle.
  2. Pack-resolved or transparent degrade. Resolve the pack; narrate pack_verified. No pack → general principles only, stated as such; never fake tailoring.
  3. Stop at every gate; never auto-publish. Every step lands decision_state: pending (architecture §9.2). The publish step hands to publish-social (its own registry-gated fork) and stops; the human publishes (D-8).
  4. Legible, not autopilot. Every step narrates its applied play + the launch's bet. Never a black box (D-8).
  5. Close the loop. End at measure-results, which writes back into the pack. No measure handoff = open loop → DONE_WITH_CONCERNS.

Quality Gate — 5 dimensions

Critic (agents/bundle-critic-agent.md) PASS/FAIL — all 5 must PASS:

  • Chain integrity — 7 steps (+4b) named; each maps to a real leaf or gate; current-step pointer present.
  • Pack legibilitypack_verified narrated; bound steps cite their pack §N; unwired steps degrade transparently.
  • Gate discipline — each step decision_state: pending; publish never auto-publishes.
  • Bundle coherence — ONE deliverable; run-of-show consistent; required artifacts present or flagged; one channel per run.
  • Loop closure — the measure-results handoff present; else flag the open loop.

Before Starting

Resume a prior bundle for this channel+slug under docs/forsvn/artifacts/marketing/launch/ (read Current step). Resolve the pack via the soft client (_shared/hosted-pack-client.md) — current pack if a hosted key serves fresher, else the local _shared/platform-intelligence/ mirror. Full: _shared/before-starting-check.md [PLAYBOOK].

Pre-Dispatch

_shared/pre-dispatch-protocol.md. Cold Start needs: channel, product/topic, brand_mode, launch goal. Mode: _shared/mode-resolver.md; --fast skips planning/revision on clean PASS, never the 5 gates or no-auto-publish. Profile: _shared/execution-policy.md.

Dispatch

Per cycle: read state → resolve the pack → dispatch the current step's leaf → stop at its gate → on approval, advance Current step, log the transition, narrate the next applied play. The copy step routes to write-launch for a launch-channel pack, write-social for a feed channel (--social forces the feed path). Step 4b dispatches write-outreach (hunter/supporter), review-gated. The unwired steps (plan, 4b, publish) degrade transparently — read the pack directly, don't rewire those leaves (S3.2/S3.3). Full sequence: the spec above.

2 agents (agents/*.md): Launch Architect (resolve pack → plan the 7-step chain mapped to leaves + per-step legibility + the bet) → Bundle Critic (the Quality Gate). Each leaf runs its own agents; this skill never inlines that work.

Artifact Contract

  • Path: docs/forsvn/artifacts/marketing/launch/[channel]-[YYYY-MM-DD]-run-[slug].md (the run record, distinct from write-launch's copy bundle). Lifecycle: pipeline.
  • Body (in order): the 7-step (+4b) run-of-show table (step · leaf · pack §N · artifact · gate/status) · Legibility block · Why this works · Critic verdict (5-row) · Anti-patterns.
  • Frontmatter per _shared/artifact-contract-template.md + pack_verified / applied_tactics. Cross-stack: the loop closes through measure-results, which owns the pack write-back + .forsvn/performance/[channel].tsv.

Chain Position

Deference: copy bundle alone → write-launch; single post → write-social; run-of-show alone → plan-campaign; generic loop → run-pipeline; post-launch read → measure-results. One channel per run.

Anti-Patterns

anti-patterns.md [ANTI-PATTERN] — fuse a step · auto-publish · fake a pack binding · black-box run · clone per channel · close without measure · multi-channel in one run.

Durable Rules (protected)

<!-- SLOW_UPDATE_START --> <!-- No pinned rules yet (populate via the slow-update workflow — references/slow-update-fence.md). --> <!-- SLOW_UPDATE_END -->

Completion Status

  • DONE — bundle assembled; 7 steps (+4b) mapped + narrated; pack resolved; current step clear; measure-results handoff present.
  • DONE_WITH_CONCERNS — a step lacks a leaf, the loop is open, the pack is stale, or a step degraded to general principles.
  • NEEDS_CONTEXT — missing channel, product/topic, brand_mode, or goal.
  • BLOCKED — no pack and the operator declines the degrade; an unresolved leaf GUARD_FAIL.

Worked Example

Product Hunt end to end, legibility narration visible: walkthrough [EXAMPLE].

What ships with it: 38 files

495.0 KB alongside SKILL.md, 3 of them executable

scripts/

Keep looking

Skills are one crate of 326,764. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.