launchlisted
Install: claude install-skill ucsandman/claude-skills
# /launch — full launch orchestration
You are orchestrating a product launch using two layers:
- **CLI (`launch/`, package `@marketing-studio/launch`)** — everything deterministic: state, validation, posting providers, research fetchers, payload rendering. Run it as `node "C:/Projects/animations/../dist/index.js" <cmd>` (or `launch <cmd>` if linked) -- the CLI lives in this package, `C:/Projects/animations/..`.
- **This session** — everything generative or credentialed-through-MCP: DashClaw infra calls, WebSearch synthesis, draft copywriting, and the approval gates. The CLI never holds DashClaw credentials and never sends email/SMS.
Honesty rule: each step below is labeled **[CLI]** or **[session]**. Don't do session work by shelling out, and don't reimplement CLI work inline.
Target `<dir>` is the project directory being launched (sibling of the engine repo).
Humans can drive the same engine visually with `launch ui` (local dashboard, same gates — see docs/ui-dashboard.md); this skill remains the autonomous path.
## Sequence
### 1. Intake [CLI]
```
launch init <dir> --domain <domain> --price <price> [--name --tagline --audience]
```
Pull flag values from the conversation. Read back `<dir>/.launch/launch.config.json` and confirm the summary with the user. Re-run with `--force` if they correct anything.
### 2. DashClaw setup [session]
1. `select_project` (or `create_project` if this product has no DashClaw project yet).
2. `create_launch` with `declared_stack` from