| name | pipeline-start |
| description | The front door to the Pipeline pack: turn a name into a tracked, worked deal. Trigger on "turn this name into a deal", "track this prospect", "help me with my pipeline", "I met someone, help me pursue them", "work this lead", "new business with <name>", "follow up with <name>", or any open-ended business-development request where the user has a person or company but no clear next step. You name who you want to pursue; the OS captures them as a tracked record, reads what it already knows, and routes you to the next move: research, a first message, a meeting brief, or a proposal. One honest disclaimer, no full onboarding, and it never sends anything for you. Routes to prospect-init, business-context-loader, meeting-prep, email-drafter, proposal-writer, list-pruner, or reconnect-prompt depending on where the deal is.
|
| why | A founder running BD alone does not think in skills, they think 'I met someone, now what'. This is the one entry that turns a name into a tracked deal and points at the right next move, so nobody has to learn which of seven pipeline skills they need first. |
| enhance | Have the person's name and company to hand, and a one-line memory of where you met or why they matter. With that the OS can open a real prospect record in the same session instead of a blank one. |
| summary | Name who you want to pursue; the OS tracks them and routes you to the next move. |
| allowed-tools | ["Read","Bash"] |
| mcp_requirements | [] |
Pipeline Start
Runs on: reasoning - this skill reads what the OS already knows and routes you to the member skill that does the work. It does not run an engine of its own and it never sends a message for you. On a read-only or cloud surface, it explains the same route without claiming a file was written.
This is the front door to your pipeline. You have a name, a company, or a lead. Underneath that one person sit several moves: capture them so they are tracked, learn what you can, prep the call, write the first message, scope the work. This skill reads what the OS holds, then points the same name at the move you actually need next, without making you learn which skill does what.
The principle: do more for you than you asked, in priority order, and never invent a fact you do not have. If the OS does not know whether they have budget, who else decides, or whether they replied, it says so rather than guessing.
The moves behind one name
| You want | What it means | Where it routes |
|---|
| Track them | get this person out of your head and into a record you can work | prospect-init (writes companies/prospects/<slug>.md) |
| Know them | what you already hold on them or their company before you reach out | business-context-loader (for a company you run) or a quick research pass |
| Reach them | a first message or a follow-up in your voice | email-drafter |
| Meet them | a brief for the call: what to ask, what to watch for | meeting-prep |
| Scope them | a proposal or quote once the conversation is real | proposal-writer |
| Clean a list | a contact export to prune before any outreach | list-pruner |
Not sure where the deal is? Say "just get them tracked" and the OS runs prospect-init first (the cheapest, most useful move), then names what each later step would add so you choose with the record in front of you.
Read the state before routing
The route is only as good as what it knows. Before choosing, read brain/.snapshot.md. If it is missing, or its date: line is more than 3 days old, run first and read the fresh one. The open-flags block tells you what is stuck, the must-do block tells you what this week is actually about, and the patterns block tells you what keeps happening - a front door that routes without them is a cold start at the exact moment the user is deciding whether this OS knows them. If Python is unavailable, route from the files you can read and say so.
