vcmatch

name	vcmatch
description	Match the active startup profile against a VC fund. Usage: /vcmatch <VC Fund Name> [profile-file.md]. Reads STARTUP_PROFILE.md by default (or the profile file if provided), spawns 4 parallel domain research sub-agents, and produces a structured VC Match Report with 8-dimension fit scoring summing to /100, portfolio pattern analysis, deal process intel, and warm intro paths.
license	MIT
compatibility	Requires Claude Code with WebSearch, WebFetch, and Agent. Internet connection required.
allowed-tools	WebSearch WebFetch Read Write Agent
metadata	{"author":"3Flux","version":"3.1","workflow-step":"4 — run after /vcposer scores 60+"}

/vcmatch — VC Match Analysis

Match the active startup profile against a VC fund using 4 parallel domain research sub-agents. Produces a 16-section report with an 8-dimension fit score summing to /100, portfolio pattern analysis, deal process intelligence, and warm intro paths.

Invocation

/vcmatch <VC Fund Name> [profile-file.md]

Argument order is flexible — if any argument ends in .md, treat it as the profile file; all other arguments form the fund name.

Examples:

• /vcmatch a16z → fund a16z, profile STARTUP_PROFILE.md
• /vcmatch Y Combinator → fund Y Combinator, profile STARTUP_PROFILE.md
• /vcmatch a16z SALES_PROFILE.md → fund a16z, profile SALES_PROFILE.md
• /vcmatch SALES_PROFILE.md Y Combinator → fund Y Combinator, profile SALES_PROFILE.md

Execution Steps

Execute these five steps in order.

---

Step 1 — Load Startup Context

Resolve the profile file:

• If the user passed an argument ending in .md, use that filename.
• Otherwise, default to STARTUP_PROFILE.md.
• If more than one .md argument is passed, stop and ask the user to pass a single profile file.
• If no fund name remains after extracting the profile file, stop and ask for a fund name.

Read the resolved profile file from the current working directory. Extract and hold in context:

• Company name and one-liner
• CEO name and key credentials (3–5 highlights)
• CTO name and key credentials (3–5 highlights)
• Sector / category
• Stage (Idea / MVP / MVBP / Revenue / Growth)
• Geography
• Target customer (specific buyer, not generic)
• Business model
• Traction (exact metrics — numbers, not descriptions)
• Raise amount and instrument (SAFE / priced seed / TBD)
• Known competitors
• Referral contacts (name, email, relationship if listed)

Derive the fund slug: lowercase the fund name, strip punctuation and spaces (Y Combinator → y-combinator, a16z → a16z).

If the profile file does not exist, stop and tell the user: "Create <profile-file> first — all /vcmatch research is personalized to your startup profile." (Use the resolved filename in the message.)

---

Step 2 — Poser Pre-Check

Before spawning any sub-agents, check for vcposer-<fund-slug>.md in the current directory.

• Score < 40: Stop. Tell the user: "This fund scored [N]/100 on vcposer — likely a poser. Run /vcposer <fund> to review the evidence, or move to the next fund."
• Score 40–59: Continue, but flag for a warning banner in the final report. Tell the user: "Proceeding with caution — this fund scored [N]/100 on vcposer."
• Score 60+ or file not found: Proceed normally.

---

Step 3 — Spawn 4 Parallel Research Sub-Agents

Read the four agent prompt files from references/:

• skills/vcmatch/references/agent-thesis.md
• skills/vcmatch/references/agent-portfolio.md
• skills/vcmatch/references/agent-people.md
• skills/vcmatch/references/agent-deal.md

Build a self-contained prompt for each sub-agent by substituting the following placeholders in the agent file text:

Placeholder	Value
{{FUND_NAME}}	The fund name as entered by the user
{{FUND_WEBSITE}}	Best-guess domain (e.g., a16z.com)
{{SECTOR}}	Startup's sector/category from profile
{{ADJACENT_SECTOR}}	1–2 closely related sectors
{{STAGE}}	Startup's current stage
{{GEOGRAPHY}}	Startup's geography
{{ONE_LINER}}	Startup's one-liner
{{RAISE_AMOUNT}}	Raise amount
{{BUSINESS_MODEL}}	Business model
{{TRACTION}}	Key traction metrics
{{COMPETITORS}}	Known competitors
{{COMPETITOR_CATEGORY}}	Generic competitor category
{{CEO_NAME}}	CEO name
{{CEO_CREDENTIALS}}	CEO key credentials
{{CTO_NAME}}	CTO name
{{CTO_CREDENTIALS}}	CTO key credentials
{{REFERRAL_CONTACTS}}	Referral names/orgs from profile
{{REFERRAL_NAME}}	Primary referral name
{{REFERRAL_ORG}}	Primary referral organization
{{INSTRUMENT}}	Raise instrument

Fire all 4 Agent tool calls in a single message so they run in parallel. Each sub-agent must receive a fully self-contained prompt string — sub-agents have no access to the parent conversation context.

Instruct each sub-agent to return only its named field blocks (as defined in the agent file). Sub-agents must not write files, score anything, or make recommendations.

Wait for all 4 sub-agents to complete before proceeding.

---

Step 4 — Synthesize and Score

Collect the returns from all 4 sub-agents. Reconcile any conflicting data points, noting the source conflict.

Score the 8 fit dimensions based on the sub-agent evidence. Each dimension must cite specific evidence — no inflated or uncalibrated scores.

Dimension	Max Points	Evidence Source
Sector Fit	20	Thesis sub-agent: Sector Clusters, De Facto Focus
Stage Fit	20	Thesis sub-agent: Stage Focus + Portfolio sub-agent: Entry Stage Reality
Thesis Alignment	20	Thesis sub-agent: Stated Mandate, De Facto Focus, Thesis Quotes
Team-Fund Fit	10	People sub-agent: Team-Fund Fit Signals
Traction Signal Fit	10	Portfolio sub-agent: Traction Bar
Check Size / Terms Fit	10	Deal sub-agent: Check Size Reality, Instrument Preference
Geographic Fit	5	Thesis sub-agent: Geography + Deal sub-agent
Network / Warm Path	5	People sub-agent: Warm Intro Paths, Referral Overlap

Overall Fit Score: Sum of all 8 dimensions. Maximum: 100.

Recommended Action thresholds:

• 75–100 → Pursue: Strong mandate fit, viable path in. Proceed to /vcintro.
• 50–74 → Warm Up First: Good fit signals but missing a proof point or intro. Address the gap, then return.
• 0–49 → No-Go: Material mandate mismatch or insufficient traction for this fund's bar. Move to the next fund.

---

Step 5 — Write Output and Report

Read the output template from references/output-format.md. Populate all 16 sections using the synthesized evidence and scores from Step 4.

Write the completed report to vcmatch-<fund-slug>.md.

The following strings must appear verbatim in the output (downstream skills vcintro, vctrack, and vcdebrief parse them by exact match):

• Recommended Action: followed by Pursue, Warm Up First, or No-Go
• Fit Score: followed by the numeric score

Then tell the user:

• The Fit Score and Recommended Action
• If Pursue: "Run /vcpartner <fund> <partner name> to profile the right GP, then /vcintro vcmatch-<fund>.md to draft outreach."
• If Warm Up First: "The gap is [state the specific dimension with the lowest score]. Address it, then re-run /vcmatch."
• If No-Go: "This fund is a poor fit — [state the primary reason]. Run /vclist to identify better-matched funds."
• "Run /vctrack to update your pipeline view."

---

Output Rules

• Every score in the Fit Assessment must cite specific evidence — not "seems like a good fit"
• The Warm Intro Path section must name specific people or companies — never generic ("ask a portfolio founder")
• The Cold Email must use [Startup Name] — never hardcode a company name
• The Verdict table must include a row labeled exactly Recommended Action
• If fewer than 3 of the 4 sub-agents return usable data, downgrade the Recommended Action by one tier (Pursue → Warm Up First, Warm Up First → No-Go) and note the data gap
• Word budget: no limit on the full report, but each individual prose section should stay under 200 words; use tables and bullets for density
• Do not fabricate portfolio company names, check sizes, or partner quotes — cite sources or mark as Unknown