company-context

name	company-context
version	2.3
last_updated	"2026-04-08T00:00:00.000Z"
author	marketing-team
description	Gathers firmographics, traction signals, funding history, team composition, tech stack, hiring activity, and key decision-makers for a target company via Exa, Apollo, Firecrawl, and Apify MCP. Produces a structured company-context document with qualification score, ICP fit assessment, red flag analysis, and optional account brief (job postings, key people, outreach angles). Triggers: "company research", "company background", "qualify this prospect", "discovery call prep", "account brief", or any company URL provided for research. Upstream: none (root context skill). Downstream: feeds competitor-research, icp-research, client-discovery, positioning, client-proposals, and apollo-sequences. NOT for competitor analysis (use /competitor-research) or ICP personas (use /icp-research).
goal	Gathers firmographics, traction signals, funding history, team composition, tech stack, hiring activity, and key decision-makers for a target company via Exa, Apollo, Firecrawl, and Apify MCP.
outcome	Gathers firmographics, traction signals, funding history, team composition, tech stack, hiring activity, and key decision-makers for a target company via Exa, Apollo, Firecrawl, and Apify MCP. Produces a structured company-context document with qualification score, ICP fit assessment, red flag...
primitive	research
ontology_type	company-context
review_gate	1
inputs	{"required":[],"recommended":[]}
outputs	[{"type":"company-context","feeds_into":["client-discovery","competitor-research","client-proposals"]}]
depends_on	[]
feeds_into	["client-discovery","client-proposals","competitor-research"]
owned_by_agent	researcher
mcps_used	["apollo-io","exa","gdrive"]
push_targets	["gdrive","notion"]
triggers	{"slash_commands":["/company-context"],"natural_language":[]}
status	draft
locked_by	null
locked_date	null
lock_version	null
sources_count	0
context	fork
effort	max
paths	marketing/**, marketing/**

Company context

Extract firmographics, traction signals, funding, team composition, tech stack, hiring activity, and decision-makers for a target company. Produces a markdown artifact with qualification score (0-25), ICP fit assessment, red flag analysis, and optional Apollo account brief. Output drops into any client or prospect folder and feeds discovery prep, competitor research, positioning, and proposal scoping.

When to run

Invoke for "company research", "company background", "qualify this prospect", "discovery call prep", "account brief", or whenever the user provides a company URL for research.

Do NOT invoke for competitor analysis (/competitor-research), product messaging extraction (/messaging), ICP personas (/icp-behavioural), or casual website checks.

Skill chain: this is a root/gateway skill. Common downstream chains in references/integration-map.md.

Inputs

Required: Company identifier — website URL, LinkedIn URL, or company name. If name is ambiguous (e.g., "Atlas", "Beam"), confirm with user before proceeding.

Optional (improve quality):

• LinkedIn company URL — sharper team size + org structure
• Specific questions — focus research on areas of interest
• Discovery call date — adds urgency context
• Account brief request — triggers Apollo enrichment (~1-2 credits, see references/account-brief.md)

Substrate: Exa-first per .claude/rules/exa-protocol.md. Primary tools company_research_exa and web_search_exa. MCP fallback chain in references/mcp-integration.md. Cite per ontology: [VERIFIED: exa_search, {url}, accessed {YYYY-MM-DD}].

Apify bulk-mode fallback (added 2026-05-01)

Imported via: /steal analysis 2026-05-01 (.claude/discovery/0526-apify-linkedin-actors-steal-analysis.md).

For ABM-scale company-context sweeps (>50 accounts in one run), Apollo's per-credit cost compounds — and Apollo doesn't index every company in the long tail. Bulk fallback:

Tool	Use case	Cost
dev_fusion/Linkedin-Company-Scraper	Bulk LinkedIn company URL → firmographics (name, industry, size, website, employee count, description, specialties)	$8/1k flat

MCP invocation pattern:

mcp__apify__call-actor
  actor: "dev_fusion/Linkedin-Company-Scraper"
  input: {
    "company_urls": ["https://www.linkedin.com/company/...", ...]
  }

Threshold rule: Use Apollo apollo_enrich_company + Exa company_research_exa for <50 companies (richer data, qualification scoring depth). Use this Apify slot for >50 companies in one ABM-scale sweep where the per-account depth needed is firmographic-only.

Concrete win: A 100-firm account brief sweep using dev_fusion at $8/1k costs $0.80, vs Apollo at ~$50 in credits (60x cheaper at this scale).

Steps

1. Validate input. Confirm company is identifiable. If name ambiguous, ask user for URL or clarification.
2. Fetch website. Pull homepage (positioning, customer signals), about page (story, team), careers page (hiring signals).
3. Search funding. Crunchbase, Tracxn, PitchBook, CB Insights, TechCrunch, Forbes, company press. Query patterns in references/search-strategies.md.
4. Search revenue / team. GetLatka, Growjo, LeadIQ, Owler, LinkedIn company page. Use revenue-from-team-size heuristics in references/search-strategies.md when sources are sparse.
5. Extract customer signals. Logo walls, case studies, "trusted by X" claims, G2/Capterra review counts.
6. Assign confidence levels. High = official source; Medium = reputable third-party; Low = aggregator/estimate. Map to [VERIFIED] / [INFERRED] / [ESTIMATED] per ontology. Detail in references/qualification-rubric.md.
7. Identify conflicting data + document gaps. Flag sources that disagree. Mark missing data as [UNAVAILABLE: searched X, Y, Z]. Suggest discovery questions to fill gaps.
8. Calculate qualification score (0-25). Five criteria × 0-5: stage fit, revenue fit, industry fit, marketing leader present, ICP signals. Rubric in references/qualification-rubric.md.
9. Assess ICP fit signals. Team signals, marketing gaps (opportunities), positioning complexity, intent signals — checklist in references/qualification-rubric.md.
10. Check red flags. Layoffs, funding drought, exec departures, pivots, hidden pricing, missing logos, high burn — checklist in references/qualification-rubric.md.
11. Generate 3-5 key observations. Discovery-call-ready insights and talking points.
12. Optional — account brief mode. If user requests, gate on Apollo credits, then run apollo_get_organization_job_postings (1 credit) + apollo_search_people (free, filtered to director+). Output hiring activity, key people, outreach angles. Spec in references/account-brief.md.
13. Self-evaluate. Run completeness, evidence, and guardrail checks per references/self-evaluation.md. Flag low-confidence areas before review gate.
14. Render output. Use the standard template in references/output-template.md (traction signals, funding details, team breakdown, qualification score, ICP fit, red flags, key observations, data gaps + optional account brief).
15. Review gate (Level 1, quick review). Present qualification score + traction signals. Then surface chain suggestions: competitors? discovery questions? client-discovery? Apollo sequence? Save as reference example if positive feedback.

What good looks like

References

• references/output-template.md — full markdown output structure + pre-delivery quality checklist
• references/qualification-rubric.md — target questions, priority sources, confidence scoring, qualification criteria, ICP fit signals, red flags, anti-hallucination guardrails
• references/account-brief.md — Apollo-credit-gated mode (job postings, key people, outreach angles)
• references/self-evaluation.md — completeness/evidence/guardrail checks + post-output iteration prompts
• references/process-flowchart.md — full visual flowchart of phases, checkpoints, review gate, chain suggestions
• references/mcp-integration.md — Exa-first substrate, MCP table (Exa, Firecrawl, Apollo, YouTube, Slack, Granola), fallback chain
• references/integration-map.md — feeds-into / receives-from skills, recommended workflow sequences, trigger phrases
• references/anti-examples.md — bad patterns, gotchas (invented data, wrong company, missing score, outdated signals, tech-stack speculation)
• references/search-strategies.md — search query templates, data source categories, revenue/team/funding heuristics
• references/skill-improvement-log.md — changelog (v1.0 → v2.3), feedback signals, reference-example capture format, pattern detection rules

Examples

Worked examples (one per company) in references/:

• example-lovable.md — $6.6B AI app builder
• example-cursor.md — $29.3B AI IDE
• example-vercel.md — $9.3B frontend cloud
• example-perplexity.md — $20B AI answer engine
• example-sierra.md — $10B AI customer service
• example-gamma.md — $2.1B AI presentations
• example-openai.md — frontier AI labs

Evaluations

• All 4 target questions answered (revenue, customers, funding, team) or marked [UNAVAILABLE: searched X]
• Every data point carries source + confidence level
• Qualification score calculated 0-25 with notes per criterion
• ICP fit + red flags checklists completed
• Data gaps section with discovery questions
• ≥3 sources per major claim, ≥50% [VERIFIED] per .claude/rules/exa-protocol.md
• No invented funding/revenue/team numbers — [UNAVAILABLE] notation when missing
• Confidence levels match the ontology mapping (High → verified, Medium → inferred, Low → estimated)

Push

After delivery, propose: /competitor-research, /discovery, /proposal, or /icp-research. If user approves output as exemplary, capture as reference example per references/skill-improvement-log.md.