@botchain/init-toolkit

Single-command bootstrap that turns Claude Code, Codex, Cursor, and Windsurf into specialized BOT Chain engineering agents. Installs:

1. A progressive skill directory (skills/botchain-skill-engine/) into every supported IDE's global skills folder.
2. A local MCP server (mcp-server/) exposing 11 tools for environment checks, contract compilation/deployment, transfers, and explorer lookups.

Install

npx github:manueldezman/botchain-init-toolkit --global

This copies the skill into ~/.claude/skills/, ~/.codex/skills/, and ~/.cursor/skills/, and registers the MCP server in each tool's config (Claude Desktop/Code, Cursor, Windsurf configs as JSON; Codex's config.toml).

Windsurf is handled differently. Instead of copying a skill folder into an unverified ~/.windsurf/skills/ path, the installer writes a condensed Rules block to Windsurf's best-effort global rules location and, when run inside a project, also writes .windsurf/rules/botchain-skill-engine.md. The installer always prints the manual fallback path: Settings → Cascade → Rules.

Other flags:

npx github:manueldezman/botchain-init-toolkit --dry-run     # show what would happen, write nothing
npx github:manueldezman/botchain-init-toolkit --skip-mcp    # install skill files only

Documentation gaps found in BOT Chain's official docs

While building this, I audited dev-docs.botchain.ai directly and found several real issues (broken links, a wrong RPC method name on the paymaster page, an invalid JSON example, contradictory faucet instructions, and more) — see FEEDBACK.md for the full list, formatted for the PR/Bug/Optimization submission track.

This toolkit's own reference docs (skills/botchain-skill-engine/references/) are written defensively around those gaps — e.g. rpc-endpoints.md and migration.md.

After installing

The installer writes blank values for the network env vars. Before using any chain tool, set them — either in your shell, or by editing the MCP entry the installer added to your tool's config:

BOTCHAIN_RPC_URL=...
BOTCHAIN_CHAIN_ID=...
BOTCHAIN_EXPLORER_API=https://scan.bohr.life
BOTCHAIN_PRIVATE_KEY=...      # testnet-only dev key. Never a mainnet key.

Restart your IDE/agent so it picks up the new skill and MCP server, then try:

"Check the BOT Chain balance for 0x..." "Deploy a simple storage contract to BOT Chain testnet"

Project layout

bin/            the npx installer (bin/install.js)
skills/         the SKILL.md + references/ + assets/ loaded by each IDE
mcp-server/     the MCP server (11 tools, see mcp-server/README.md)

See mcp-server/README.md for the tool list and how to run/test the server standalone, and skills/botchain-skill-engine/SKILL.md for how the agent is instructed to use it.

Safety notes

• execute_terminal_cmd only runs an allow-listed set of binaries (forge, npx, npm, which, git) via execFile with an argv array — there is no shell string parsing, so there's no shell-injection surface.
• Every side-effecting tool (deploy_contract, send_transaction, batch_transfer, and anything that installs software) is documented in SKILL.md as requiring the user's explicit approval in the conversation. The server itself doesn't gate on that — the calling agent is expected to ask first, same as any other MCP tool.
• The toolkit targets the BOT Chain testnet by default. Nothing points at mainnet unless you configure it to.

License

MIT-0 — free to use, modify, and redistribute. No attribution required.