| name | raise-match |
| description | Use to deeply analyse whether a specific VC fund fits your startup — spawns four parallel research sub-agents (thesis, portfolio, people, deal), scores the fit across eight dimensions out of 100, and writes a complete fund dossier with the warm-intro path, perks, and a pursue/warm-up/no-go verdict. Run after raise-vet clears the fund. Use when asked to "analyse this fund", "does this VC fit us", "research a fund", or "should we pitch this investor". |
| allowed-tools | WebSearch WebFetch Read Write Agent |
| metadata | {"track":"fundraising"} |
raise-match — the gated, four-agent fund dossier
The workhorse of the track. Where raise-vet asks "is this fund real?",
raise-match asks "does it fit us?" — and answers it with real research:
four sub-agents in parallel, an evidence-cited fit score, and a single dossier the
rest of the track builds on. It is the sole assembler of dossier.md.
Invocation
raise-match <fund name>
Steps
1 — Load context and the vet gate
Read .spine/raise/profile.md (stop → raise-init if absent). Derive the fund
slug. Read .spine/raise/funds/<slug>/dossier.md if it exists and branch on
legitimacy_score (see ADR 0017):
- < 40 → stop: "This fund scored [N]/100 on vet — likely a poser. Run
raise-vet <fund> to review, or move to the next fund." Spawn nothing.
- 40–59 → proceed, and flag a caution banner in the dossier.
- ≥ 60, or no vet yet → proceed normally (suggest the founder run
raise-vet
first if there's no dossier, but don't block).
2 — Fan out four research sub-agents
Read the four prompt templates and fill the profile placeholders
({{FUND_NAME}}, {{SECTOR}}, {{STAGE}}, {{GEOGRAPHY}}, {{ONE_LINER}},
{{TRACTION}}, {{RAISE_AMOUNT}}, {{BUSINESS_MODEL}}, {{CEO_NAME}},
{{CEO_CREDENTIALS}}, {{CTO_NAME}}, {{CTO_CREDENTIALS}},
{{REFERRAL_CONTACTS}}, {{COMPETITORS}}):
- references/agent-thesis.md — mandate, focus, stage, geography, quotes
- references/agent-portfolio.md — recent deals, entry-stage reality, traction bar
- references/agent-people.md — partners, team-fund fit, warm-intro paths
- references/agent-deal.md — check size, instrument, process, perks
Fire all four Agent calls in a single message so they run in parallel. Each
prompt must be fully self-contained — sub-agents see none of this
conversation. Instruct each to return only its named field blocks: no scoring, no
recommendations, no file writes. Wait for all four.
3 — Synthesise and score
Reconcile conflicting data (note the conflict and its sources). Score eight fit
dimensions, each citing specific evidence from the returns — never an
uncalibrated guess:
| Dimension | Max | Evidence |
|---|
| Sector fit | 20 | thesis: focus clusters |
| Stage fit | 20 | thesis: stage + portfolio: entry-stage reality |
| Thesis alignment | 20 | thesis: mandate + quotes |
| Team–fund fit | 10 | people: fit signals |
| Traction signal fit | 10 | portfolio: traction bar |
| Check size / terms | 10 | deal: check size + instrument |
| Geographic fit | 5 | thesis/deal: geography |
| Network / warm path | 5 | people: warm-intro paths |
Fit score = the sum (max 100). Recommended action: ≥75 → pursue ·
50–74 → warm-up · <50 → no-go. If fewer than three sub-agents returned usable
data, downgrade the action one tier and note the gap.
4 — Write the dossier (whole)
Rewrite .spine/raise/funds/<slug>/dossier.md end to end using
references/dossier-format.md. Preserve the
## Legitimacy section if raise-vet wrote one, and complete the handoff
header (fit_score, recommended_action, updated). Then report the fit score
and verdict, and the next step: pursue → raise-pitch / raise-outreach;
warm-up → name the lowest-scoring dimension to fix; no-go → state the primary
mismatch and suggest raise-funds.
Spine I/O
- Reads:
.spine/raise/profile.md; .spine/raise/funds/<slug>/dossier.md (vet stub).
- Writes:
.spine/raise/funds/<slug>/dossier.md (full dossier; sole assembler).
Notes
- All scoring lives here, not in the sub-agents — so calibration is in one
place and scores stay comparable across funds.
- Never fabricate portfolio names, check sizes, or partner quotes — cite sources
or mark
Unknown. Each prose section stays under ~200 words; lead with tables.
- This is the only skill that writes the full
dossier.md (ADR 0016).