SKILL.md
Proposal Builder
The proposal's job is to make the decision easy. One specific paid next step with a clear outcome, timeline, and price. A menu of services is a request for the buyer to do your thinking; they will do it by not replying.
Workflow
- Collect the call notes. Required: the problem in the prospect's words, company name, budget range, timeline, and who decides. Missing any of these: ask once before drafting, a proposal aimed at the wrong decision-maker or budget tier is wasted work.
- Pick the smallest credible scope. Match the engagement to certainty, not to revenue hope:
- Diagnostic (fixed small fee): they need a risk map, an assessment, or a prioritized fix list before committing bigger. Also the default when budget, timeline, or authority is unclear.
- Fixed-scope project (fixed fee, 1-3 weeks): one stuck workflow or system that can be stabilized fast.
- Retainer: recurring operations, maintenance, reporting, enablement, or multiple workflows. When in doubt, propose the diagnostic. A small closed deal beats a large stalled one.
- Write the eight sections, one page total:
- Situation: what they said is broken or at risk, in their words
- Business cost: time, missed revenue, support load, risk, delay
- Scope: 3-5 concrete deliverables, no vague "strategy"
- Out of scope: kills ambiguity before it kills the project
- Timeline: start date, milestones, decision deadline if one really exists
- Price: fixed fee or retainer amount, payment terms
- Proof: brief credibility tied to implementation, not hype
- Next step: one action that approves the work
- Run the quality bar before sending: names the real failure mode, scope small enough to close now, buyer effort low, a concrete shipped artifact is promised, the CTA is a yes/no approval. Cut anything that sounds like a generic consultancy.
- Save it as markdown in your pipeline folder, e.g.
$WORKDIR/pipeline/proposals/<company>-<date>.md, and log the follow-up date.
Verification
Read the finished proposal and answer from the buyer's chair: "Do I know exactly what I get, by when, for how much, and what I do to say yes?" Expect four crisp answers pulled straight from the page. If any answer requires inference, that section is vague: rewrite it before sending. Then check length. Expect one page. Over: cut proof and situation first, never scope clarity.
Good vs Bad
Bad: "We offer AI strategy, implementation, training, and ongoing support. Packages start at $X/month. Let us know which offering interests you and we can schedule a call to discuss options."
Good: "Your order-sync automation fails silently about twice a week and ops spends Fridays reconciling (your estimate: 6 hrs/wk). Two-week stabilization: root-cause fix, retry-and-alert layer, runbook, handoff session. $<FIXED_FEE>, half on start. Starts Monday the <DATE> if approved by Thursday. Reply 'approved' and I'll send the invoice and kickoff invite."
Completion checklist
- [ ] Problem, budget, timeline, decision-maker captured from the call
- [ ] One scope proposed, sized to certainty, not a menu
- [ ] All eight sections present, one page total
- [ ] Out-of-scope section included
- [ ] Deliverables concrete, each one a shippable artifact
- [ ] CTA is a single yes/no approval action
- [ ] Buyer-chair test passed
Any box unchecked: not done. Fix or say so.
Footguns
- The menu proposal. Three options with feature grids feels generous, reads as indecision, and stalls the deal. Fix: one recommendation; mention at most one fallback in a single line.
- Scope creep in the proposal itself. Every extra deliverable adds a reason to say "let me think about it". Fix: 3-5 deliverables max, everything else goes to out-of-scope or a future phase.
- Fake deadlines. An invented "prices go up Friday" torches trust with exactly the buyers you want. Fix: use real constraints (calendar slots, project start windows) or state none.
- Bio padding. Long credentials sections signal insecurity. Fix: two or three proof lines tied to work like theirs, that is all.