| name | sales-research |
| description | Company Research & Firmographic Analysis Subagent — evaluates Company Fit (25% of Prospect Score) across 8 research dimensions and 5 scoring sub-dimensions using structured web intelligence. Trigger standalone via `/sales research <url>` producing COMPANY-RESEARCH.md, or as subagent 1 during `/sales prospect <url>` returning Company Fit Score 0-100. Industry-agnostic engine.
|
| weight | 0.25 |
| tier | spoke |
Company Research & Firmographic Analysis
Tier 0 Doctrine
- Tier: Spoke (subagent of sales-prospect orchestrator; also standalone via
/sales research <url>)
- Authority: Reads discovery briefing from orchestrator or raw URL from user; scores company fit only
- Determinism first: All scoring uses declared sub-dimension formulas and calibration tables. No LLM interpretation of scores.
- No fabrication: Every data point must have a source. Absence of data is scored, not invented. Revenue estimates must state methodology and confidence.
IMO (Top-Level)
| Layer | Responsibility |
|---|
| Ingress | Company URL + optional discovery briefing with pre-fetched pages (schema validation only) |
| Middle | 8-dimension web research; 5-sub-dimension scoring with evidence; strength/risk/insight synthesis |
| Egress | COMPANY-RESEARCH.md (standalone) or Company Fit Score 0-100 with structured data (subagent) — read-only output |
Constants
| Constant | Value | Authority |
|---|
| Research Dimensions | 8 fixed (see Workflow) | Locked |
| Scoring Sub-Dimensions | Size Fit, Industry Fit, Growth Trajectory, Tech Sophistication, Budget Signals | Fixed; 5 sub-dimensions |
| Sub-Dimension Range | 0-20 each | Fixed |
| Score Formula | sum of 5 sub-dimensions | Produces 0-100 |
| Calibration: 16-20 | Exceptional — clear evidence, ideal range, multiple confirming sources | Fixed |
| Calibration: 11-15 | Strong — good evidence from 2+ sources, within acceptable range | Fixed |
| Calibration: 6-10 | Moderate — some signals, partially fits criteria | Fixed |
| Calibration: 1-5 | Weak — limited signals, marginal fit | Fixed |
| Calibration: 0 | Disqualifying — evidence actively contradicts fit | Fixed |
| Source Priority (8 levels) | 1. Company website, 2. SEC/public filings, 3. Crunchbase/PitchBook, 4. LinkedIn, 5. Press releases, 6. News articles, 7. Review sites (G2/Capterra/Glassdoor), 8. Social media | Fixed hierarchy; higher wins on conflict |
| Revenue Estimation Methods | Employee-based ($200K-$300K/employee), Funding-based (A=$1-3M, B=$5-15M, C=$15-50M ARR), Customer-based (count x avg tier), Traffic-based (traffic x CVR x AOV) | Fixed; must state method + confidence |
| Confidence Levels | High, Medium, Low, Speculative | Fixed |
| Data Freshness: Employees | Within 6 months; flag if older | Fixed |
| Data Freshness: Funding | Must include most recent round; flag if 18+ months stale | Fixed |
| Data Freshness: News | Last 6 months for Recent Developments; older goes to History | Fixed |
| Tech Stack Signal Sources | Job postings, Website source, Integration pages, Developer docs, Blog posts, Conference talks | Fixed; 6 sources |
| Web Search Queries | 7 structured queries per company (see Block 2) | Fixed pattern |
| Output: Standalone | COMPANY-RESEARCH.md | Fixed |
Variables
| Variable | Source | Runtime |
|---|
target_url | User input or orchestrator briefing | Per-run |
company_name | Homepage detection | Discovered |
discovery_briefing | Orchestrator (subagent mode only) | Per-run |
invocation_mode | standalone or subagent | Per-run |
homepage_content | WebFetch of target URL | Discovered |
interior_pages | WebFetch of up to 9 key pages (about, team, pricing, blog, careers, customers, press, legal, contact) | Discovered |
tech_stack_signals | 6 signal sources (job posts, source code, integrations, dev docs, blog, talks) | Discovered |
search_results | 7 WebSearch queries | Discovered |
company_overview_data | Dimension 1 extraction | Discovered |
business_model_data | Dimension 2 extraction | Discovered |
product_tech_data | Dimension 3 extraction | Discovered |
leadership_data | Dimension 4 extraction | Discovered |
funding_data | Dimension 5 extraction | Discovered |
market_position_data | Dimension 6 extraction | Discovered |
culture_data | Dimension 7 extraction | Discovered |
recent_dev_data | Dimension 8 extraction | Discovered |
size_fit_score | Calibration table | Computed |
industry_fit_score | ICP match assessment | Computed |
|
Workflow
BLOCK 1: Website Intelligence Collection
Governed by: C&V
Constants: 9 page types (about, team, pricing, blog, careers, customers, press, legal, contact), 6 tech stack signal sources, source priority hierarchy
Variables: target_url, homepage_content, interior_pages, tech_stack_signals, discovery_briefing
IMO:
- Ingress: URL validated as reachable; if discovery briefing present, pre-fetched pages loaded; skip already-fetched pages
- Middle:
- Fetch homepage via WebFetch — extract company name, tagline, value prop, product positioning, social proof
- Fetch up to 9 interior pages (About, Team, Pricing, Blog, Careers, Customers, Press, Legal, Contact) — skip any provided in briefing
- For each page: store URL, title, raw content, key data points
- Detect tech stack from 6 signal sources: job postings (required skills), website source (meta tags, scripts, framework signatures), integration pages (listed partners), developer docs (API tech, SDKs), blog posts (technical content), conference talks (architectural choices)
- If URL unreachable: attempt www/non-www and https/http variants; if still unreachable, report error
- If specific page not found: note "Not publicly available", proceed with available data
- Egress: Structured page content store + tech stack inventory
Go/No-Go: Proceed if homepage accessible. If zero pages accessible, halt and report URL error to user.
BLOCK 2: External Research & 8-Dimension Extraction
Governed by: IMO
Constants: 7 web search query patterns, 8 research dimensions, data freshness rules, source priority hierarchy, revenue estimation methods
Variables: search_results, company_overview_data through recent_dev_data
IMO:
- Ingress: Company name + all page content from Block 1
- Middle:
- Execute 7 WebSearch queries:
"[company]" overview, "[company]" funding round, "[company]" revenue employees, "[company]" CEO founder, "[company]" news recent, "[company]" reviews Glassdoor, "[company]" competitors market
- Resolve conflicting data using source priority hierarchy (company website > SEC > Crunchbase > LinkedIn > press > news > reviews > social)
- Apply data freshness rules: flag employee data older than 6 months; flag funding older than 18 months; note revenue estimation methodology and confidence
- Extract data for 8 dimensions:
- Dim 1 — Company Overview: Name, founded, founders, HQ, offices, employee count, stage, mission, vision, structure
- Dim 2 — Business Model & Revenue: Revenue model, pricing tiers, revenue estimate (use estimation methods constant), customer count, key metrics, unit economics
- Dim 3 — Product & Technology: Core products, category, tech stack, differentiators, roadmap signals, integrations, API/platform, patents, open source
- Dim 4 — Leadership & Team: CEO/founder, CTO, key executives, board, advisory, recent changes, public presence, leadership style
- Dim 5 — Funding & Financial Health: Total funding, latest round, round history, key investors, valuation, burn rate signals, profitability path
- Dim 6 — Market Position: Market category, competitors (top 3-5), market share estimate, competitive advantages, win/loss signals, analyst coverage, awards
- Dim 7 — Culture & Employer Brand: Values, Glassdoor rating + themes, hiring pace, work model, DEI signals, benefits, employer brand strength
- Dim 8 — Recent Developments (6 months): Product launches, partnerships, funding events, leadership changes, market moves, controversies, customer wins, acquisitions
- Egress: 8 dimension data objects with source citations per data point
Go/No-Go: Proceed unconditionally. If web search returns limited results, note data gap and reduce confidence. Always extract from whatever is available.
BLOCK 3: 5-Sub-Dimension Scoring
Governed by: CTB
Constants: 5 sub-dimensions (Size Fit, Industry Fit, Growth Trajectory, Tech Sophistication, Budget Signals), calibration scale 0-20 per sub-dimension, score formula
Variables: size_fit_score, industry_fit_score, growth_trajectory_score, tech_sophistication_score, budget_signals_score, company_fit_score
IMO:
- Ingress: All 8 dimension data objects from Block 2
- Middle:
- Size Fit (0-20): Score by employee range calibration: 1-10 (5-10), 11-50 (10-15), 51-200 (15-20), 201-1000 (12-18), 1001-5000 (8-15), 5000+ (5-12). Adjust within range based on trajectory (growing vs stable vs declining).
- Industry Fit (0-20): Score by ICP alignment: exact match (15-20), adjacent with relevance (10-14), some relevance (5-9), minimal relevance (1-4), mismatch (0). If no ICP available, score based on general engagement signals.
- Growth Trajectory (0-20): Score by growth signals: rapid hiring 20%+ in 6mo (15-20), recent funding <6mo (12-18), new launches/expansion (10-15), steady 5-15% growth (8-12), stable flat (3-7), declining/layoffs (0-3).
- Tech Sophistication (0-20): Score by tech maturity: modern/API-first/developer-focused (15-20), modern SaaS tools + integrations (10-14), standard with some modern (5-9), legacy/limited (1-4).
- Budget Signals (0-20): Score by evidence: enterprise pricing/"Contact Sales" (15-20), recent funding Series B+ (12-18), hiring for roles using product category (10-15), multiple paid tools in stack (8-12), bootstrap/price-sensitive (2-6), clear budget constraints (0-2).
- Validate each sub-score is 0-20 integer
- Compute
company_fit_score = size_fit + industry_fit + growth_trajectory + tech_sophistication + budget_signals
- Egress: Score breakdown table with per-sub-dimension evidence
Go/No-Go: Output is always produced. Any sub-dimension scored 0 must include explicit data-gap note explaining why.
BLOCK 4: Synthesis & Output Assembly
Governed by: Circle
Constants: Strengths format (3-5, statement + evidence + sales implication), Risks format (3-5, statement + evidence + mitigation), Key Insights format (5, non-obvious + actionable + sourced + recommendation), output schema
Variables: strengths, risks, key_insights, invocation_mode, company_fit_score
IMO:
- Ingress: All dimension data + all sub-dimension scores from Blocks 2-3
- Middle:
- Compile Strengths (3-5): each with specific evidence, source citation, and sales implication
- Compile Risks (3-5): each with specific evidence, source citation, and mitigation strategy
- Extract Key Insights (5): each must be non-obvious (not learnable in 30 seconds from homepage), actionable (informs sales approach), sourced, with recommendation
- Write Executive Summary: 2-3 paragraphs covering who they are, what they do, trajectory, fit assessment
- If
invocation_mode = standalone: assemble full COMPANY-RESEARCH.md (see Output Template pointer)
- If
invocation_mode = subagent: return structured data block with Company Fit Score, sub-dimension breakdown, company snapshot fields, top strengths, top risks
- Terminal display (standalone): condensed summary with Unicode bar charts (10-char bars, filled=U+2588, empty=U+2591)
- Egress: COMPANY-RESEARCH.md written to disk (standalone) or structured score block returned (subagent)
Go/No-Go: Output is always produced. Clearly note all data gaps. If multiple dimensions have no data, set overall confidence to Low and recommend manual research.
Output Template
# Company Research: [Company Name]
**URL:** [url]
**Date:** [current date]
**Company Type:** [type]
**Industry:** [vertical]
**Company Fit Score: [X]/100**
---
## Executive Summary
[2-3 paragraphs: who they are, what they do, trajectory, fit assessment]
## Company Snapshot
| Field | Value |
|-------|-------|
| **Company Name** | [name] |
| **Founded** | [year] |
| **Founders** | [names] |
| **Headquarters** | [location] |
| **Employees** | [count] (source: [source]) |
| **Stage** | [Startup/Growth/Mature/Public] |
| **Total Funding** | [amount] |
| **Latest Round** | [round type, amount, date] |
| **Revenue Estimate** | [range] (method: [method], confidence: [H/M/L/S]) |
| **Key Investors** | [names] |
| **Tech Stack** | [key technologies] |
## 1. Company Overview
## 2. Business Model & Revenue
## 3. Product & Technology
## 4. Leadership & Team
## 5. Funding & Financial Health
## 6. Market Position
## 7. Culture & Employer Brand
## 8. Recent Developments
## Company Fit Score: [X]/100
| Sub-Dimension | Score | Evidence |
|--------------|-------|----------|
| Size Fit | [X]/20 | [key evidence] |
| Industry Fit | [X]/20 | [key evidence] |
| Growth Trajectory | [X]/20 | [key evidence] |
| Tech Sophistication | [X]/20 | [key evidence] |
| Budget Signals | [X]/20 | [key evidence] |
| **Total** | **[X]/100** | |
## Strengths
1. **[Strength]** — [Evidence]. *Sales implication: [how to use]*
## Risks
1. **[Risk]** — [Evidence].
— [Evidence].
Terminal Output (Standalone Mode)
=== COMPANY RESEARCH COMPLETE ===
Company: [name] ([type])
Industry: [vertical]
Stage: [stage]
Employees: [count]
Funding: [total]
Revenue Est.: [range]
Company Fit Score: [X]/100
Size Fit: [XX]/20 ████████░░
Industry Fit: [XX]/20 ██████░░░░
Growth Trajectory: [XX]/20 ███████░░░
Tech Sophistication:[XX]/20 █████░░░░░
Budget Signals: [XX]/20 ████████░░
Top Strengths:
1. [strength]
2. [strength]
3. [strength]
Top Risks:
1. [risk]
2. [risk]
Full report saved to: COMPANY-RESEARCH.md
Rules
- Never invent data points. Every fact requires a source citation. "They probably have X" is not evidence.
- Never score optimistically when data is absent. Unknown = score at midpoint of sub-dimension range, not top.
- Never omit the estimation methodology for revenue figures. State the method and confidence level for every estimate.
- Never use employee count data older than 6 months without flagging staleness.
- Never count funding rounds older than 18 months as "recent" for growth trajectory scoring.
- Never treat a single review or social post as a signal. Patterns across sources are signals; isolated mentions are noise.
- Never hardcode industry names in scoring logic. Industry fit is scored against ICP context (a variable), not a constant vertical list.
- Never skip a research dimension. If data is unavailable for a dimension, report "Not publicly available" and note the gap.
Reference Pointers
| Reference | Location |
|---|
| Orchestrator | skills/sales-prospect/SKILL.md |
| ICP definition | IDEAL-CUSTOMER-PROFILE.md (working directory, optional) |
| Decision maker skill | skills/sales-contacts/SKILL.md |
| Qualification skill | skills/sales-qualify/SKILL.md |
| Competitive intel skill | skills/sales-competitors/SKILL.md |
| Doctrine | templates/doctrine/ARCHITECTURE.md (IMO, Hub-Spoke, CTB) |
| Skill creation rules | skills/skill-creator/SKILL.md |
