Sales Call Prep

Sales Call Prep

Deterministic helpers pull CRM history + web pages; you, the agent, reconstruct the timeline, deep-dive company + person, map product-to-prospect, and write the call strategy. The prior-interaction line is the single most call-changing fact — always check CRM + outreach before web research.

v2.0.0 — depth upgrade. Adds brief_scaffold.py — a deterministic pre-call scaffold that classifies the prospect's authority tier + buying role, infers a communication style + adaptation guidance, predicts the top objections with reframe stubs, and emits the full brief-section skeleton. This SKILL.md carries the authority matrix, the communication-style matrix, the predicted-objection logic, and the brief skeleton.

When to use

• "Prep me for my sales call", "call prep for [company]", "research [person] before our demo".
• Distinct from meeting-brief: this is "how do I sell to this person".

How to run

Step 1 — prior interactions (deterministic CRM read + agent)

python3 ${SKILL_DIR}/scripts/crm_history.py --crm hubspot \
  --company "Acme" --domain acme.com --output ${WORKSPACE}/history.json

Pulls the company's associated deals + contacts (HubSpot or Pipedrive) and sets has_prior_contact. For Salesforce/Close or an outreach tool (lemlist/instantly campaign, emails sent, their reply, lead category), read those APIs directly (agent-routed) and merge. Build a chronological timeline; never re-pitch someone who already replied; acknowledge lost deals; multi-thread off colleagues. If no CRM/outreach is configured, state "true cold meeting" explicitly and run pure web research.

Step 2 — company deep dive, sales lens (deterministic fetch + agent)

python3 ${SKILL_DIR}/scripts/fetch_url.py --url https://acme.com \
  --output ${WORKSPACE}/company.json

Across 7 dimensions: overview, financial/growth signals, pain/need indicators, tech/competitive stack (mine job postings via your search tool), decision landscape, relationship map, recent news (90d). Crunchbase/LinkedIn/job boards are JS/auth — route via web-automation + Robomotion Proxy.

Step 3 — person deep dive (deterministic scaffold + agent)

python3 ${SKILL_DIR}/scripts/brief_scaffold.py \
  --title "VP of Engineering" --company "Acme" --deal-stage evaluation \
  --has-prior-contact false --competitor "Incumbent X" --price-tier premium \
  --output ${WORKSPACE}/scaffold.json

Emits a deterministic scaffold.json: authority classification, communication style + adaptation, predicted objections (ranked, with reframe stubs), and the brief skeleton. Then add the researched substance — professional profile, what they care about (posts/content). LinkedIn is anti-bot — prefer phantombuster (session cookie) or web-automation + proxy; degrade to search snippets.

Authority-classification matrix (title cue → buying role → engagement strategy):

Communication-style adaptation (role family → style → how to adapt):

Predicted-objection logic — the scaffold ranks objections from context (competitor in play → "we already use X"; premium price → "too expensive"; technical title → integration/ security; IC/manager → "not the decision-maker"; mid/late stage → "need to think"; cold → "why are you reaching out"; always → status-quo). Each ships a reframe stub; you turn it into acknowledge → reframe → proof, with a same-industry/size proof point.

Step 4 — product-to-prospect mapping + strategy (you, the agent)

Connect genuine pain → your capability + proof (don't force-fit; never trash competitors). Proof-point priority: same industry > same size > same problem > same competitor > general. Write the full brief using the scaffold's brief_skeleton as the section list:

Outputs

• A single pre-call document with all sections above. Workspace + Agent Teams channel attachment.

Credentials / env

• Required: none for the core — fetch_url.py and brief_scaffold.py are keyless and the strategy synthesis is your job as the agent (no LLM key in the script layer).
• Optional: if a CRM key is set (HUBSPOT_API_KEY / PIPEDRIVE_API_TOKEN, or Salesforce/Close) → crm_history.py reconstructs the prior-interaction timeline; if not → keyless CSV/paste of past touches, else "cold meeting" + web research (default). If an outreach key is set (LEMLIST_API_KEY / INSTANTLY_API_KEY) → outreach history; if not → skip. If PHANTOMBUSTER_API_KEY + LinkedIn cookie are set → reliable person research; if not → keyless serp/fetch (default). APIFY_API_TOKEN — if set → their-product review scraping; if not → keyless fetch. Google Calendar MCP for calendar-driven mode.

Notes & edge cases

• Always check CRM + outreach before web research — the prior-interaction line changes the call.
• Map only genuine pain→capability connections; differentiate on approach, don't trash competitors.
• crm_history.py matches by --domain first (more reliable), then --company name.