Sales Training Landscape Research

The Claude Code ecosystem is rich for sales execution (outreach, prospecting, call analysis) but nearly empty for sales training: nothing existing does audible role-play, difficulty ladders, scorecard trends, or fear-of-selling coaching. That gap is exactly what the SalesTraining project should fill.

The plan: install two skill packs as knowledge, adopt Founding Sales + SPIN as the operating methodology (with Hormozi's CLOSER for one-call closes and April Dunford's pitch structure as the spine), run graduated-exposure fear work based on the call-reluctance research, and build eight custom skills including a voice-ready /roleplay gym. Day one requires zero build: Claude's mobile voice mode already handles spoken drills.

Top repos

• louisblythe/Sales-Skills (83 stars): 122 skills in three buckets, including 20 Core Sales Skills (active listening, SPIN-style questioning, rapport, discovery, objection handling, negotiation, closing, storytelling). The Core 20 are the best knowledge base to feed a role-play grader. Plugin-ready.
• coreyhaines31 sales-enablement skill: builds objection quick-reference cards, talk tracks, and role-play scenario docs. Same family as skills already installed (offers, positioning-basics). Closest existing thing to a training skill.
• zubair-trabzada/ai-sales-team-claude: 14 skills + 5 parallel agents for prospect research, BANT/MEDDIC qualification, outreach, meeting prep, proposals. Execution-focused, useful for the "then actually go sell" half.
• sales-skills/sales: broad GTM utility pack (CRM, outbound, enrichment). Not training.
• OneWave-AI/claude-skills (172 skills) and alirezarezvani cold-email (330+ skills): big grab-bags, worth cherry-picking with a security review each, not bulk-installing.
• northrn/claude-sales-skills (lead gen, smaller) and anthropics/skills (nothing sales-training specific, good for SKILL.md conventions).

Marketplaces and roundups

The ClaudSkills sales category lists roughly 1,984 sales-tagged skills plus a cold-outreach use-case page; virtually all are outreach and CRM automation, not rep training. Other roundups: MCP Market, SyncGTM, AI Builder Club, ClaudeCodeHQ, Salesforge, Novoslo.

One pattern worth stealing from the AI Builder Club stack: a sales-context skill that writes one context file (ICP, value prop, proof points) which every other sales skill reads. Direct analog to the project's _context/ pattern; every role-play persona and scorecard should read a per-business offer-context.md.

What the commercial "sales gyms" do

• Hyperbound (see also the AgentRank review): voice-first AI buyer bots trained on 2M+ hours of real B2B calls; a custom buyer bot matching your exact ICP takes about 10 minutes. AI Scorecards grade talk ratio, key selling moments, methodology adherence, and coaching areas instantly after every session.
• Second Nature: visual avatars, 4 practice formats, and manager-assigned moods per persona. Its critique (role-plays feel too easy for experienced reps) is why difficulty calibration matters.
• Roundups: Kendo AI, Mindtickle, Nooks 2026 comparison, Federico Presicci, ELM Learning, AI Tools Bakery.

Core mechanics across all of them, every one replicable in Claude Code with markdown personas plus a grading pass: (1) persona bot built from the ICP, (2) mood and difficulty dial, (3) scenario library (cold call, discovery, demo, objection gauntlet, negotiation), (4) instant scorecard with quotes from the transcript, (5) repeat-the-same-hard-moment reps, (6) trend tracking over time.

Prompt patterns that work

• A good role-play prompt has 5 elements: buyer persona, scenario context, behavioral instructions (skeptical, busy, friendly), objection type, and conversation structure. Sources: Auto Interview AI prompt engineering guide and their top 10 prompts.
• Difficulty ladder: Level 1 warm and curious → Level 2 cautious, needs approval, raises 2 objections → Level 3 satisfied with their current solution, no budget, only took the call as a favor. Objections should compound in sequence, advancing only when the previous one is handled.
• Scoring: rate each weighted criterion 0-100 with a transcript quote justifying every score; critical criteria can fail the call outright. Dimensions: opening hook, question quality, objection handling, closing ask. Sources: Sybill on objection role-play, Sybill's 10 prompts, Close: stress-test your script, PitchMonster, Pitchbase's 12 scenarios.
• Key realism rule: keep the AI buyer terse and human (short answers, interruptions, "I only have 5 minutes"), never helpful. The buyer and the coach/grader must be separate turns or separate agents so the buyer never breaks character.

Voice options for audible role-play

1. Claude mobile voice mode (recommended first path, zero build). Live on iOS, Android, and web with hands-free and push-to-talk modes; Anthropic explicitly positions it for practicing important conversations. Paste a persona prompt into a Claude project and run spoken drills on walks. Sources: Claude Help, Tom's Guide, Android Headlines, TechCrunch.
2. Local laptop pipeline for scored, logged sessions: microphone → Whisper (local STT) → Claude (buyer persona + grader) → edge-tts (free, natural voices) → speaker. Reference builds: real-time voice agent with Whisper + VAD, local-talking-llm, Speech-to-Speech-Local-LLM, DeepInfra guide. Different edge-tts voices per persona (gruff medspa owner vs. chirpy gym manager) is free realism.
3. Claude Code's own voice mode exists, but the mobile app is the better daily-rep surface; the laptop pipeline is the better scored-and-archived surface since transcripts land in the vault for review.

The field

• SPIN (Situation, Problem, Implication, Need-payoff): best for discovery-heavy consultative selling where buyers haven't fully understood their problem. HubSpot's 12 methodologies, SPOTIO's comparison.
• Sandler: mutual-qualification posture, uncovers pain, budget, and decision up front, and removes the "chasing" dynamic. Great for fear: you're evaluating them too. Business.com guide.
• Challenger: teach, tailor, take control with a provocative insight; powerful but a heavy lift for small teams and can feel aggressive for a warm, relationship-led seller. Salesmotion, IRC Sales Solutions.
• Gap Selling: current state vs. future state; suits deals where the buyer underrates the problem (very true of medspas and trainers who don't know what AI ops could do for them).
• Hormozi CLOSER: Clarify → Label → Overview past attempts → Sell the vacation → Explain away concerns → Reinforce. Built for one-call closes on high-ticket services, low-pressure, extremely scriptable. Sales Coach Pro guide, Hormozi's own thread, Skool explainer, complete guide.
• Founding Sales (Pete Kazanjy): THE founder-led sales handbook, free online at foundingsales.com (table of contents, Amazon, First Round podcast, founder-led sales pattern, GTMnow playbook). Kazanjy's framing is tailor-made for the fear: modern sales is consulting where the consultant has a predilection for one solution, theirs. Founders are uniquely equipped because they genuinely understand the problem.
• April Dunford's Sales Pitch: 8-component pitch structure. Setup (market insight → honest pros and cons of the alternatives, including "do nothing" as the biggest competitor) then Follow-through (differentiated value → proof → objections → the ask). Pairs perfectly with the positioning skills already installed. Amazon, Lenny's Newsletter walkthrough, summary.

• The canonical text: The Psychology of Sales Call Reluctance (Dudley & Goodson) (Internet Archive copy, expanded edition). Thirty years of research; call reluctance is "the fear of self-promotion," broken into 12 distinct types (Over-Preparer, Yielder who fears intrusion, Social Self-Consciousness, Role Rejection: shame about being "in sales"). Each type has matched countermeasures: Threat Desensitization (graduated exposure) and Thought Realignment (cognitive restructuring). Key implication: diagnose the type first (likely Role Rejection + Yielder for a warm relationship-oriented founder), then pick drills. Wrong-drill-for-type is why generic "just make more calls" advice fails.
• Rejection therapy: Jia Jiang's 100 Days of Rejection (TED talk, Rejection Proof book) is exposure therapy applied to asking. Two findings transfer directly: desensitization is fast when reps are daily and low-stakes, and asking "why?" after a no often converts it or teaches something.
• Practical drills (Close: 11 ways to overcome call reluctance): time-boxed call blocks, the first-call-of-the-day rule (scariest one first), script the first 20 seconds so the amygdala never gets a vote, count activities not outcomes (a "no" scores the same point as a "yes"), and reframe identity from "selling to" to "helping and screening."

• 30/60/90 structure: the standard onboarding framework. First 30 days: product knowledge assessments, methodology certification, observed calls, with 100% of calls reviewed days 1-15 tapering to 25% by day 90.
• Practice volume: a typical new rep runs 20-30 practice conversations in the first two months; AI role-play lets you repeat the same hard moment ten times in an afternoon. Teams running structured role-play report roughly 37% performance lift and 30% faster ramp. Hyperbound, FullyRamped, Enboarder.
• Certification gates: give the learner the scorecard ahead of time, let them practice against it, then a graded certification call gates progression. Mindtickle, WorkRamp, Highspot.
• Film review: record everything, review with a rubric, extract one improvement per review. PitchMonster's 12 scenarios with scripts.

The 30-day program sketch

Skills to build (all read a shared per-business offer-context file)

Install vs. build

• Install (as knowledge, security-reviewed first): the louisblythe Core 20 (objection-handling, effective questions, discovery, closing, active listening) and the coreyhaines sales-enablement skill. The offer and positioning side is already covered by installed skills (offers, positioning-basics, obviously-awesome, market-research, strategic-planning).
• Maybe later (execution phase): ai-sales-team-claude for prospecting-to-proposal once real calls are happening, wired to the GBP Prospect Finder pipeline.
• Build ourselves: everything in the table above. Nothing in the ecosystem does audible role-play, difficulty ladders, scorecard trends, or fear-of-selling coaching.