/company-research

name: company-research
description: |
  Company discovery and deep research skill. Researches a company's product and ICP,
  discovers target companies to sell to using Browserbase Search API, deeply researches
  each using a Plan→Research→Synthesize pattern, and scores ICP fit — compiled into
  a scored research report and CSV. Supports depth modes (quick/deep/deeper) for
  balancing scale vs intelligence.
  Use when the user wants to: (1) find companies to sell to, (2) research potential
  customers, (3) discover companies matching an ICP, (4) build a target company list,
  (5) do market research on prospects. Triggers: "find companies to sell to",
  "company research", "find prospects", "ICP research", "target companies",
  "who should we sell to", "market research", "lead research", "prospect list".
license: MIT
compatibility: Requires browse CLI (`npm install -g browse`) and BROWSERBASE_API_KEY env var
allowed-tools: Bash Agent
metadata:
  author: browserbase
  version: "1.1.0"

Company Research

Discover and deeply research companies to sell to. Uses Browserbase Search API for discovery and a Plan→Research→Synthesize pattern for deep enrichment — outputting a scored research report and CSV.

**Required**: `BROWSERBASE_API_KEY` env var and `browse` CLI installed.

**First-run setup**: On the first run you'll be prompted to approve `browse cloud fetch`, `browse cloud search`, `cat`, `mkdir`, `sed`, etc. Select **"Yes, and don't ask again for: browse cloud fetch:\*"** (or equivalent) for each to auto-approve for the session. To permanently approve, add these to your `~/.claude/settings.json` under `permissions.allow`:

"Bash(browse:*)", "Bash(bunx:*)", "Bash(bun:*)", "Bash(node:*)",
"Bash(cat:*)", "Bash(mkdir:*)", "Bash(sed:*)", "Bash(head:*)", "Bash(tr:*)", "Bash(rm:*)"

**Path rules**: Always use the full literal path in all Bash commands — NOT `~` or `$HOME` (both trigger "shell expansion syntax" approval prompts). Resolve the home directory once and use it everywhere. When constructing subagent prompts, replace `{SKILL_DIR}` with the full literal path.

**Output directory**: All research output goes to `~/Desktop/{company_slug}_research_{YYYY-MM-DD}/`. This directory contains one `.md` file per researched company plus a final `.csv`. The user gets both the scored spreadsheet and the full research files on their Desktop.

**CRITICAL — Tool restrictions (applies to main agent AND all subagents)**:

• All web searches: use `browse cloud search`. NEVER use WebSearch.
• All page content extraction: use `node {SKILL_DIR}/scripts/extract_page.mjs "<url>"`. This script fetches via `browse cloud fetch --output`, parses title + meta tags + visible body text, and automatically falls back to `browse get markdown` when fetch fails or returns thin JS-rendered content. NEVER hand-roll a `browse cloud fetch | sed` pipeline — it strips meta tags and doesn't parse the stdout JSON envelope. NEVER use WebFetch.
• All research output: subagents write **one markdown file per company** to `{OUTPUT_DIR}/{company-slug}.md` using bash heredoc. NEVER use the Write tool or `python3 -c`. See `references/example-research.md` for the file format.
• Report + CSV compilation: use `node {SKILL_DIR}/scripts/compile_report.mjs {OUTPUT_DIR} --open` — generates HTML report and CSV in one step, opens overview in browser.
• URL deduplication: use `node {SKILL_DIR}/scripts/list_urls.mjs /tmp` after discovery.
• **Subagents must use ONLY the Bash tool. No other tools allowed.**
• **Main agent NEVER reads raw discovery JSON batch files.** Use `list_urls.mjs` for dedup.

**CRITICAL — Anti-hallucination rules (applies to main agent AND all subagents)**:

• NEVER infer `product_description`, `industry`, or `target_audience` from a site's fonts, framework (Framer/Next.js/React), design system, or typography. These are cosmetic and say nothing about what the company sells.
• NEVER let the user's own ICP leak into a target's description. If you don't know what the target does, write `Unknown` — do not pattern-match them onto the ICP.
• `product_description` MUST quote or paraphrase a specific phrase from `extract_page.mjs` output (TITLE, META_DESCRIPTION, OG_DESCRIPTION, HEADINGS, or BODY). If none of those fields yield a recognizable product statement, write `Unknown — homepage content not accessible`.
• If `product_description` is `Unknown`, cap `icp_fit_score` at 3 and set `icp_fit_reasoning` to `Insufficient evidence — homepage returned no readable content`.

**CRITICAL — Minimize permission prompts**:

• Subagents MUST batch ALL file writes into a SINGLE Bash call using chained heredocs. One Bash call = one permission prompt.
• Batch ALL searches and ALL fetches into single Bash calls using `&&` chaining.

Pipeline Overview

Follow these 5 steps in order. Do not skip steps or reorder.

1. **Company Research** — Deeply understand the user's company, product, and who they sell to 2. **Depth Mode Selection** — Choose research depth based on how many targets they want 3. **Discovery** — Find target companies using diverse search queries 4. **Deep Research & Scoring** — Research each company, score ICP fit 5. **Report & CSV** — Present findings, compile scored CSV

---

Step 0: Setup Output Directory

Before starting, create the output directory on the user's Desktop:

OUTPUT_DIR=~/Desktop/{company_slug}_research_{YYYY-MM-DD}
mkdir -p "$OUTPUT_DIR"

Replace `{company_slug}` with the user's company name (lowercase, hyphenated) and `{YYYY-MM-DD}` with today's date. Pass `{OUTPUT_DIR}` (as a full literal path, not with `~`) to all subagent prompts so they write research files there.

Also clean up discovery batch files from prior runs:

rm -f /tmp/company_discovery_batch_*.json

Step 1: Deep Company Research

This is the most important step. The quality of everything downstream depends on deeply understanding the user's company.

1. Ask the user for their company name or URL

2. **C