AI controls
When each AI step runs, the skills it uses in order, the model each runs on, and how often they ran. What a skill tells the model is on the Skills page.
Models run through OpenRouterRuns are counted from Launch's plans and each campaign's Activity
Skills with the same order run together, and each order sees the answers of the one before.
Launch
A teammate asks Launch to plan and size a campaign
| Order | Skill | Status | Model | Runs, 30 days | Failed or missing | Last run |
|---|---|---|---|---|---|---|
| 1 | Campaign plannerVersion 2026.10.01.1 | Draft | anthropic/claude-opus-5.5Any provider · Starting model | 0 | 0 | Never |
| 2 | Prospecting searchVersion 2026.10.01.1 | Draft | anthropic/claude-sonnet-5.5Any provider · Starting model | 0 | 0 | Never |
Strategy
A teammate changes who a campaign reaches
| Order | Skill | Status | Model | Runs, 30 days | Failed or missing | Last run |
|---|---|---|---|---|---|---|
| 1 | Prospecting searchVersion 2026.10.01.1 | Draft | anthropic/claude-sonnet-5.5Any provider · Starting model | 0 | 0 | Never |
| 2 | Prospect fitVersion 2026.10.01.1 | Draft | deepseek/deepseek-v4.1-flashAny provider · Starting model | 0 | 0 | Never |
Pull
A teammate asks to pull or top up a test batch
| Order | Skill | Status | Model | Runs, 30 days | Failed or missing | Last run |
|---|---|---|---|---|---|---|
| 1 | Prospect fitVersion 2026.10.01.1 | Draft | deepseek/deepseek-v4.1-flashAny provider · Starting model | 0 | 0 | Never |
A pull brings people onto a campaign
| Order | Skill | Status | Model | Runs, 30 days | Failed or missing | Last run |
|---|---|---|---|---|---|---|
| 1 | Name cleanupVersion 2026.10.01.2 | Draft | deepseek/deepseek-v4.1-flashPrivate providers only · Starting model | 0 | 0 | Never |
Copy
A teammate asks for a campaign's copy
| Order | Skill | Status | Model | Runs, 30 days | Failed or missing | Last run |
|---|---|---|---|---|---|---|
| 1 | Cold email copyVersion 2026.10.01.2 | Draft | anthropic/claude-opus-5.5Any provider · Starting model | 0 | 0 | Never |
| 2 | Campaign QAVersion 2026.10.01.2 | Draft | openai/gpt-5.6-solAny provider · Starting model | 0 | 0 | Never |
A teammate asks for the copy to be checked again
| Order | Skill | Status | Model | Runs, 30 days | Failed or missing | Last run |
|---|---|---|---|---|---|---|
| 1 | Campaign QAVersion 2026.10.01.2 | Draft | openai/gpt-5.6-solAny provider · Starting model | 0 | 0 | Never |
A teammate asks to spin approved copy
| Order | Skill | Status | Model | Runs, 30 days | Failed or missing | Last run |
|---|---|---|---|---|---|---|
| 1 | Smartlead spintaxVersion 2026.10.01.1 | Draft | anthropic/claude-sonnet-5.5Any provider · Starting model | 0 | 0 | Never |
Inbox
A reply to a hub campaign is found
| Order | Skill | Status | Model | Runs, 30 days | Failed or missing | Last run |
|---|---|---|---|---|---|---|
| 2 | Draft replyVersion 2026.09.29.3 | Draft | anthropic/claude-sonnet-5.5Private providers only · Starting model | 0 | 0 | Never |
| 1 | Reply intentVersion 2026.09.30.1 | Draft | deepseek/deepseek-v4.1-flashPrivate providers only · Starting model | 0 | 0 | Never |
| 1 | Reply sentimentVersion 2026.09.29.1 | Draft | deepseek/deepseek-v4.1-flashPrivate providers only · Starting model | 0 | 0 | Never |
A teammate asks for a new draft of an answer
| Order | Skill | Status | Model | Runs, 30 days | Failed or missing | Last run |
|---|---|---|---|---|---|---|
| 1 | Draft replyVersion 2026.09.29.3 | Draft | anthropic/claude-sonnet-5.5Private providers only · Starting model | 0 | 0 | Never |
Skills no step runs yet
Written and kept by the desk, but nothing sets them off.
| Skill | Job | Status |
|---|---|---|
Cold call scriptWrite the cold call script for an Outreach Hub campaign from its plan: an opener that names what this audience has, the offer in a few spoken lines, one ask, answers to the five objections they are most likely to raise, and a two-line note for the Salesfinity notes field. Use when a campaign has calls as a channel and an offer, in the hub or in Claude Code. Writes the script only: never dials, never writes to Salesfinity or HubSpot, and never promises a number the plan does not state. | Calling | Draft |
Enrichment routingDecide, for each person in an Outreach Hub campaign's batch, which contact details the desk asks enrichment-ops for (work email, personal email, mobile phone), which it already has or skips, and which bad values it sends back to be looked up again. Use before the desk's Enrich step, in the hub or in Claude Code, once the batch is pulled and the person has chosen what to enrich. Decides only: never calls enrichment-ops or a provider, never picks a provider or its order, never spends. | Enrichment | Draft |
Campaign verdictRead one Outreach Hub campaign's test numbers against its own kill and scale rules and recommend scale, keep testing or kill, or say there is not enough data yet, in one sentence that quotes the numbers behind it. Also says which rule lines were crossed and which could not be judged. Use each morning and whenever a test's numbers change. Recommends only: never changes a stage, sends, spends or makes the call, which is always a person's. | Reporting | Draft |
Daily digestWrite the team's morning digest for Outreach Hub: one line per active campaign with its sent, replies, connects and verdict, then the positive replies that still need an answer, then the campaigns waiting on a scale or kill decision. Answers with sections the desk can render and the same digest as plain text. Use when the desk asks for the morning summary. Writes content only: never posts, sends or decides, never forms its own verdict, and never reads a provider or a database. | Reporting | Draft |