Account Intel Brief

account-intel-brief produces a pre-call technical brief in chat. It is the first skill most SEs reach for before a discovery call, demo prep, or technical qualification. It uses grounded research when Firecrawl is available and stays assumption-based when it is not.

What this skill is

A conversational pre-call research workflow. You hand it an account seed and it returns:

• A short summary.
• A company snapshot.
• Stack signals.
• Likely initiatives.
• A discovery question bank.
• Citations, or a clearly labelled set of assumptions when grounding was not possible.

Under the hood it can call Firecrawl search, scrape, and crawl, and it can push the result to Google Docs or Confluence when you explicitly ask for writeback.

Trigger conditions

Use this skill when the user asks for:

• An account brief.
• Pre-call prep.
• Company research before discovery.
• Likely stack signals or discovery questions for an account.

Intake order

Codex asks one question at a time, in this order:

1. accountSeed
2. notes.product — what you are selling in.
3. notes.sellerNotes — any seller context, or the target stakeholder.

If accountSeed is already in $account-intel-brief ..., the first question is skipped.

Output shape

The artifact is returned in chat with these sections:

1. Summary
2. Company snapshot
3. Stack signals
4. Likely initiatives
5. Discovery questions
6. Citations or assumptions

How to best use this skill

Run it before every pre-call. The brief is cheap and returns in chat, so there is no reason to wing the first meeting.

Provide a domain if you have one. If the seed looks like a domain, Codex treats it as account.domain and grounds research against that specific site plus related sources. Plain text names work, but a domain gives you tighter citations.

Name the product you are selling in. The brief tailors stack signals and discovery questions to the product angle. Without it, you get a generic company snapshot.

Pass seller context as notes.sellerNotes. Useful examples: “Security review is deferred”, “Interested in a controlled POC”, “Champion is the VP of RevOps”. This context is woven into the discovery question bank so you do not re-ask things the account already told you.

Check grounding before quoting. Every claim in the brief is either cited or labelled as an assumption. Treat the assumptions section as a TODO list of things to confirm on the call.

Save and writeback are opt-in. The brief comes back in chat. If you want the markdown saved or pushed to Google Docs or Confluence, say so — Codex will confirm the destination and verify the toolkit is connected.

How to combine with other skills

The brief is intentionally the first skill in the canonical pre-sales chain. Every other skill can reuse it.

• Into Demo Scenario Builder. Once the brief exists in chat, the demo builder reuses the company snapshot, stack signals, and likely initiatives as context instead of re-researching. Run the brief first, then say “build a demo for this account selling Solutioneer to a RevOps lead”.
• Into Integration Fit Gap Analyzer. Stack signals from the brief are a useful starting point for the fit-gap’s target systems. If the brief surfaces “Segment + Stripe + Snowflake”, you can feed those directly into the fit-gap skill.
• Into POC Handoff Orchestrator. The POC handoff explicitly consumes prior artifacts. Keep the brief in the same chat (or reference it in artifactRefs) so the handoff package picks up company context without redoing the work.

Example invocations

Explicit:

$account-intel-brief outtake.ai

Natural language:

Prepare an account brief for Outtake.ai. We are selling Solutioneer to their RevOps lead.

Example normalized input (for reference — you never type this directly):

{
  "account": { "name": "Acme Health", "domain": "acme.example" },
  "objective": "Prepare a pre-call technical brief",
  "notes": {
    "product": "Solutioneer",
    "sellerNotes": ["Interested in a controlled POC", "Security review is deferred"]
  },
  "artifactRefs": [],
  "destinations": {},
  "providerMode": "auto"
}

Provider behavior

• With FIRECRAWL_API_KEY — grounded research via tools/firecrawl/search.ts, tools/firecrawl/scrape.ts, and tools/firecrawl/crawl.ts. Claims carry citations.
• Without FIRECRAWL_API_KEY — the brief stays explicitly local-only. Unsupported claims are labelled as assumptions. Codex does not invent external facts.
• External writeback — requires COMPOSIO_API_KEY and a connected toolkit. Supported toolkits: googlesuper (Google Docs) and confluence. See Connect Google Docs & Sheets and Connect Confluence.