Designs the fundraise before any artifact exists: whether and when to raise, the narrative and value proposition, the round's size, price, and instrument, and the terms to accept or fight. Distilled from a 625-source fundraising corpus (Paul Graham's essays, YC, NFX, CRV, Carta/Cooley data, first-hand founder and investor accounts); numbers carry source and vintage because market norms drift fast.

= the directory containing this SKILL.md. Every path below is relative to ; expand to an absolute path before reading when the CWD is elsewhere. All bundled files are reference (read-only).

references/what-vcs-evaluate.md

references/narrative-playbook.md

references/round-design.md

references/stage-benchmarks.md

references/deal-terms.md

references/devtools-positioning.md

Devtools / AI-infra companies: whichever branch runs, read in full before producing output — layer naming, the rent-duration answer, the platform-incumbent answer, and the OSS→commercial narrative change what every other reference recommends.

Step 1: Map the company. Collect from the founder and their materials: maturity stage in evidence terms (team / product / repeatable sale / unit economics — not just the round label); traction facts with time context; market and ICP; team and what's in-house; cash, runway, burn; existing cap table including every SAFE at its cap; what they think they're raising and why. Done when: every item is filled in and each SAFE's conversion at cap is modeled.

Step 2: Diagnose fundability. Read in full. Run the fundability self-assessment and the fund-math arithmetic for the intended investor class (check ≈ 2% of fund size; ownership 10–20%; a credible fund-returning exit). Deliver one of three verdicts: raise now / fix first (with the specific failing criteria and what closes them) / different capital (angels, revenue, smaller round — when venture math doesn't fit). Done when: the verdict is delivered with every failing criterion named and the fund-math computed for at least one named fund profile.

Step 3: Construct the narrative. Read in full. Produce the narrative memo: archetype and point-of-view template chosen consciously, a why-now that survives its test, the compression ladder (one-liner through two-sentence + example), the earned secret, and the risks section written by the founder's side. Done when: the memo passes every item of the narrative stress-test checklist.

Step 4: Design the round. Read in full and pull current bars from . Produce the round spec: amount (milestone-first, cross-checked against burn-first and runway-first — convergence within ~20%); instrument (via the decision matrix in ); valuation posture (a defensible range, sanity-checked with the ⅓ rule against a realistic next round); dilution budget; and the milestone map that buys the next round's bar. Done when: the round spec passes the raise-now-or-wait readiness test and the sizing worksheet is complete with all three methods shown.

Step 5: Deliver the strategy. Assemble one document: verdict + narrative memo + round spec + milestone map + the champion-arming kit (the three sentences, the one-pager, the pre-answered risks) + open gaps stated plainly. Route deck rendering to vc-pitch-deck, pipeline and meetings to vc-outreach, and term-sheet review back here when paper arrives. Done when: the strategy document is delivered and every unresolved risk or evidence gap is listed in it explicitly.

When a claim needs source-level evidence beyond these references and exists in the workspace, search the corpus with run from the directory (always ; hybrid and vector modes hang). Compiled articles live at .