Startup Sales Simulator

Practice every startup sales call before it happens.

Startup Sales Simulator is an open-source Codex skill for generating sales use cases and practicing realistic, stateful buyer conversations. Give it your startup URL, pitch, pricing page, product brief, or previous call report. It first creates a native Markdown pack of relevant call ideas, then Codex becomes any selected buyer, remembers what you said, challenges unsupported claims, and decides whether you earned the next step.

After the call, it generates a native Markdown performance report that opens directly inside Codex—no dashboard or HTML required.

What it does

• Generates 8–15 startup-specific call scenarios instead of forcing one default call.
• Creates connected multi-call gauntlets across champions, technical buyers, procurement, and decision-makers.
• Grounds every exercise in verified product, pricing, customer, and proof facts.
• Creates a plausible buyer, decision process, current alternative, risks, and objections.
• Runs the call interactively without coaching or exposing the answer key.
• Adapts buyer behavior to your questions, claims, proof, and mistakes.
• Tracks discoveries, objections, buying signals, commitments, and unsupported claims.
• Allows a real win, follow-up, loss, no decision, or responsible disqualification.
• Scores eight sales competencies from transcript evidence.
• Generates a full Markdown report with clickable links to exact call turns.
• Suggests three next calls after every report and creates retry drills from the weakest branch.
• Works naturally with Codex Voice Mode.

Install

npx --yes startup-sales-simulator@latest

Or install directly from GitHub:

npx --yes skills add Kappaemme-git/startup-sales-simulator -g -y --skill simulate-startup-sales --agent codex

Restart Codex after installation so the skill list refreshes.

Generate your use cases

Use $simulate-startup-sales to analyze https://my-startup.example and create a native Markdown pack with 12 different sales calls I can practice. Include video-ready hooks and multi-call gauntlets, but keep buyer motives and exact objections hidden.

Open the generated file directly inside Codex, choose a scenario ID, and start the call.

Start a specific call

Use $simulate-startup-sales to train me to sell https://my-startup.example.
Run a hard pricing negotiation with an agency COO.
Stay in character and do not coach me until the call ends.

Or ask for three recommendations:

Use $simulate-startup-sales to give me three very different calls I should practice for my startup.

The skill inspects the available context and recommends distinct scenarios before starting. Say start immediately when you do not want the chooser.

Training formats

| Format | Output | | --- | --- | | scenario-pack | 8–15 call ideas and 2–4 connected gauntlets in one Codex-native Markdown file | | single-call | One live buyer role-play with its own persistent session and report | | gauntlet | 3–5 separate stakeholder calls around the same deal | | retry-drill | A shorter call that attacks one weakness using changed wording and context | | transcript-review | Evidence-based scoring of an existing call transcript |

Example use cases

• The sixty-second cold call
• The twenty-minute category skeptic
• The free spreadsheet pricing objection
• The enterprise security gate
• The procurement discount squeeze
• The champion without authority
• The competitor loyalist
• The demo hijacker with fifteen feature requests
• The polite “send me information” stall
• The attractive but wrong-fit enterprise lead
• The angry former trial
• The weak-adoption renewal

Call modes

| Mode | What it tests | | --- | --- | | cold-outreach | Relevance, permission, curiosity, and a respectful exit | | discovery | Problem depth, impact, urgency, fit, and qualification | | product-demo | Relevance, tailored explanation, proof, and restraint | | pricing-negotiation | Value defense, scope, tradeoffs, and commercial discipline | | technical-buyer | Feasibility, integration, security, and honest unknown handling | | procurement-gauntlet | Vendor risk, legal constraints, consistency, and patience | | competitor-comparison | Differentiation, switching value, and evidence | | closing-call | Decision clarity and a mutual next step | | renewal-expansion | Adoption, outcomes, recovery, renewal, and expansion judgment | | angry-prospect | Listening, accountability, boundaries, and recovery |

Difficulty levels: guided, realistic, hard, and expert.

The live-call rules

During the role-play, Codex:

• speaks only as the buyer;
• produces one concise prospect turn at a time;
• does not score, praise, explain, or offer response choices;
• reveals guarded information only when you earn it;
• notices contradictions and asks for proof;
• remembers every material statement;
• does not let persistence substitute for product fit;
• ends when the conversation reaches a real decision point.

You can explicitly say pause for one coaching observation, then resume the same session.

Codex-native report

Completed calls create:

sales-simulator/
└── sessions/
    └── startup-call-session.json

outputs/
├── startup-sales-scenario-pack.md
└── startup-sales-call-report.md

The scenario pack contains call ideas, opening lines, seller missions, training focus, video hooks, and connected programs without exposing the hidden answer key.

Each completed-call report contains:

1. Simulated call outcome
2. Weighted scorecard
3. Strengths and deal blockers
4. Buying signals
5. Objection analysis
6. Claims and evidence audit
7. Missed discovery opportunities
8. Turning points
9. Better responses
10. Recommended next steps
11. Retry plan
12. Full transcript with turn links

Open the .md file directly in Codex and continue with:

Make me retry the weakest objection from this report.

Scoring

| Competency | Weight | | --- | ---: | | Discovery | 20% | | Qualification | 15% | | Clarity | 10% | | Value articulation | 15% | | Objection handling | 15% | | Evidence integrity | 10% | | Listening and adaptability | 10% | | Next step | 5% |

Every score cites exact transcript turns. Outcome and skill remain separate: a seller can win a weak call because the buyer was already committed, or score highly by responsibly disqualifying a bad fit.

Try the demo

The included OrbitDesk demo is fictional. It generates 12 calls, four multi-call gauntlets, and a difficult pricing negotiation in which the seller earns a follow-up but invents a SOC 2 claim and gets penalized for it.

npm run demo

Then open both:

outputs/orbitdesk-sales-call-report.md
outputs/orbitdesk-sales-scenario-pack.md

See DEMO.md for the video flow and reusable prompts.

Test

npm test

Tests cover scenario variety, hidden seed protection, gauntlet references, configuration validation, alternating turns, persistent state, weighted scoring, broken transcript references, Markdown generation, and installer configuration.

Integrity

The simulator does not train fake testimonials, invented compliance, deceptive urgency, manipulation, or pressure after a responsible no. Real-company context can ground a call, but fictional buyer behavior is always a simulation—not a prediction of a real person or company.

License

MIT. See LICENSE.

100% open source.