run-launch

Installation
SKILL.md

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:

Installs
1
GitHub Stars
12
First Seen
12 days ago
run-launch — hungv47/meta-skills