AI tool installation and use
The tools below are optional interfaces to the same KTAPEOUT public workflow. Choose one assistant, install it using its official documentation, and then apply the KTAPEOUT Application Engineering Skill. Do not install assistants inside the KTAPEOUT source tree or commit their caches.
Common setup
- Clone or unpack the reviewed KTAPEOUT source.
- Open the repository root in the selected assistant.
- Read
docs/index.md, the relevant protocol page and the AI skill. - Ask for a plan before allowing edits.
- Keep credentials and wallet software outside the workspace.
- Review the diff and run the repository's documented checks yourself.
The public implementation is Rust-based. The baseline local check is:
cargo check --manifest-path core/rust-cli/Cargo.toml --locked
Use an external CARGO_TARGET_DIR when building locally. The command does not grant signing or broadcast authority.
OpenAI Codex
Install and authenticate through the official Codex CLI documentation. Open the repository directory, paste the safe starter prompt from the AI skill, and ask Codex to plan before editing. Keep approval or sandbox controls enabled for commands that write files or access the network. Review every proposed protocol or dependency change before accepting it.
Claude Code
Follow the official Claude Code documentation for installation and authentication. Start Claude Code from the repository root, provide the KTAPEOUT skill as project instructions, and request a read-only plan first. Keep signing, wallet and deployment actions outside the assistant session.
GitHub Copilot
Enable Copilot using the official GitHub setup documentation. In the editor, add the KTAPEOUT skill text to the repository's reviewed AI instructions or paste it into the chat context. Ask Copilot to explain the affected schema and vector before generating code. Require human review for all protocol-facing changes and pull requests.
Cursor
Install Cursor from its official documentation, open the repository as a project, and add the KTAPEOUT skill to the project's reviewed rules or chat context. Use planning mode for protocol changes and keep network, terminal and file permissions limited to the project boundary.
Windsurf
Follow the official Windsurf documentation for installation. Open only the reviewed workspace, add the KTAPEOUT skill to the project guidance, and use its planning/review flow before accepting edits. Do not place wallet files, private keys or deployment credentials in the workspace indexed by the assistant.
Tool-neutral project instruction
If a tool supports a project instruction file, include a reviewed copy of the following short rule and link to this document:
Use the included KTAPEOUT 0.5.0 specification, schemas and local CLI guidance.
The optional complete Application Engineering Skill requires the separate engineering distribution.
Preserve protocol identifiers, hashes, bounds and vectors. Never access secrets,
wallets, private keys, signing, broadcasting or credentialed endpoints. Plan
before editing, validate against schemas and run the documented checks.
Tool installation, model access and account permissions are managed by each tool vendor. KTAPEOUT does not require or endorse a specific AI provider.