# CT Web Solutions — AI Builder Starter Kit
First edition preview • September 14, 2026

## Your outcome
Build one small, understandable agent workflow that reads supplied material, produces a structured result, and stops for review. Then explore the architecture of a prediction-market research assistant. This edition contains a learning guide and an offline workflow example; it does not contain CT’s original Kalshi bot or a live trading integration.

## 1. Start with a bounded job
Choose one task: summarize three supplied project notes into a checklist. Write the input, output, success condition, allowed tools, maximum steps, and actions requiring approval before using an agent. Start with local sample files rather than accounts or credentials.

Success condition: every checklist item points back to a note; unsupported claims are marked unknown. Stop after one draft and one review. Keep a small set of examples so you can check behavior after changing prompts or tools.

## 2. Choose your workspace
Hermes is a third-party agent project from Nous Research. Follow its official installation and quickstart documentation. CT links to it; CT does not own Hermes. Model/provider costs and system requirements depend on your chosen setup. Review those before installing.
- https://hermes-agent.nousresearch.com/desktop
- https://hermes-agent.nousresearch.com/docs/

For coding assistants, start with a single relevant skill from CT’s existing Design Ops kit. Read its contents and permissions before installing. The five listed skills are Frontend Design Ops, Dashboard Design Ops, Landing Page Design Ops, Codex Review Ops, and Clip Editing Ops. They are alternatives for different jobs, not five prerequisites.
- https://ctwebsolutions.com/resources/design-ops-kit
- https://code.claude.com/docs/en/skills

Keep API keys out of prompts, screenshots, source files, and shared archives. Use your provider’s supported credential mechanism and configure a spending limit where available. Installing a skill does not make its instructions trustworthy automatically.

## 3. Run the included offline example
Prerequisite: Python 3.10 or later. No packages, accounts, keys, or paid services are required for this example.
Run `python3 workflow_demo.py`. It reads the included fictional market snapshots and writes a research checklist to standard output. Run `python3 -m unittest test_workflow.py` to check validation behavior.

This is a deterministic workflow scaffold, not an LLM agent or trading strategy. The stages are ingest → validate → report → human review. Later, an optional model can explain validated records, but it must not replace validation or invent missing market facts.

Try changing a fixture probability to 1.2, deleting a source, or changing the timestamp. Invalid input should fail, not quietly become an analysis. A real integration also needs current source retrieval, staleness checks, retries, rate limits, authentication where required, and recorded provenance.

## 4. Kalshi build path
CT’s audience has asked about the Kalshi bot. The educational product should teach the build in stages:
1. Read market information and preserve source, timestamp, and contract rules.
2. Validate data; distinguish quoted prices from estimated probabilities.
3. Produce a research note with known facts, assumptions, and unanswered questions.
4. Add a separate simulation with explicit assumptions for fees, spreads, liquidity, and fills.
5. Test an appropriate documented demo environment before evaluating any production integration.

The current included example only implements stages of data validation and research reporting with fictional records. It does not simulate fills, estimate profit, connect to Kalshi, or place orders. CT’s original video source code has not been reviewed or included. A tutorial should not repeat profitability claims without substantiated evidence.

Kalshi’s demo uses mock funds and separate credentials; demo behavior can differ from real markets. Official reference: https://docs.kalshi.com/getting_started/demo_env

## 5. Perpetuals are a separate advanced track
Kalshi already documents perpetual futures. Predictions and perpetuals have separate API references; do not reuse an event-contract implementation as though it were a perpetuals integration. Perpetuals require margin access approval and introduce leverage, funding, and liquidation considerations. This edition links to official learning material; no perpetuals implementation is included.
- https://docs.kalshi.com/welcome
- https://help.kalshi.com/en/articles/15357656-applying-for-perpetuals-access
- https://help.kalshi.com/en/articles/15357671-risks-of-trading-perpetual-futures

## Troubleshooting
Python missing: install a supported version from python.org and reopen your terminal.
File missing: run the command from the extracted kit folder.
Validation error: check the field named in the error against the fictional fixture; do not bypass validation.
Agent makes unsupported claims: reduce the task, provide source text, require traceable evidence, and reject unsupported output.
Agent repeats indefinitely: enforce maximum steps and stop conditions outside the model.

## Edition and support boundaries
This is a local product preview, not a released paid course. External software, services, and pricing can change. Links were checked during preparation on September 14, 2026. No ongoing support or lifetime updates are promised. Third-party software is linked rather than redistributed. The downloadable archive contains CT-branded draft learning material and an original offline example only.