Imported from ThinkOffApp/ide-agent-kit (SKILL.md). Install upstream with npx skills add ThinkOffApp/ide-agent-kit. Copyright stays with the author.

IDE Agent Kit

Connect your IDE coding agents into real-time teams through OpenClaw. Filesystem-based message bus, room polling, automation rules, and multi-model agent coordination. Zero dependencies.

Security Model

This skill operates in two tiers:

Core (local-only, no credentials needed):

Local filesystem queue and receipt log — agents read/write files in the working directory
init, receipt tail, memory (local backend), keepalive — no network, no secrets
serve binds to 127.0.0.1 only by default — receives webhooks, writes to local queue

Advanced (requires explicit opt-in and credentials):

sessions, gateway — talk to an OpenClaw gateway (requires openclaw.token in config)
poll — connects to Ant Farm rooms (requires --api-key flag)
emit, hooks create — POST data to external URLs you specify
tmux run, exec — execute shell commands (restricted to an allowlist in config)

No advanced features activate without explicit configuration. The default init config has empty credential fields and a minimal command allowlist.

Network behavior

Command	Outbound connections	Inbound connections
`init`, `receipt tail`, `memory` (local), `keepalive`	None	None
`serve`	None	localhost:8787 only (configurable)
`poll`	Ant Farm API (HTTPS)	None
`sessions`, `gateway`	OpenClaw gateway (localhost by default)	None
`emit`	User-specified URL	None
`hooks create`	User-specified webhook URL	None

Command execution

tmux run and exec only run commands listed in tmux.allow in your config. Default allowlist: npm test, npm run build, pytest, git status, git diff. Commands not on the list are rejected.

exec adds an approval flow: commands go through exec request → human/agent exec resolve before running.

Quick Start

npm install -g ide-agent-kit
ide-agent-kit init --ide claude-code

Creates a local ide-agent-kit.json config. All credential fields are blank. Nothing connects to any server until you configure it.

Connectivity Modes

Four modes that compose freely. Only mode 1 is active by default.

1. Local Filesystem Bus (default)

Agents on the same machine communicate through a shared queue directory and receipt log. No network, no server, no API keys.

Queue: ./ide-agent-queue.jsonl
Receipts: ./ide-agent-receipts.jsonl

2. Webhook Relay Server (optional)

Receives inbound webhooks from GitHub/GitLab and writes them to the local event queue.

ide-agent-kit serve [--config <path>]

Binds to 127.0.0.1:8787 by default. Set github.webhook_secret in config to verify signatures. Does not make outbound connections.

3. Ant Farm Room Polling (optional)

Connects to Ant Farm rooms for cross-machine coordination.

ide-agent-kit poll --rooms <room1,room2> --api-key <key> --handle <@handle> [--interval <sec>]

Requires: --api-key flag (Ant Farm API key). Rate-limited, default 120s interval.

4. GitHub Events (optional)

When serve is running, point a GitHub webhook at your relay URL. Translates PR/issue/CI events into local queue events.

Requires: github.webhook_secret in config to verify inbound signatures.

Commands

Core (local-only, no credentials)

Command	Description
`init [--ide <name>] [--profile <balanced\|low-friction>]`	Generate starter config
`receipt tail [--n <count>]`	Print last N receipts
`watch [--config <path>]`	Watch event queue, nudge IDE session on new events
`serve [--config <path>]`	Start webhook relay server (localhost only)
`memory list\|get\|set\|search`	Manage agent memory (local file backend)
`keepalive start\|stop\|status`	Prevent macOS sleep for remote sessions

Advanced (requires credentials or explicit config)

Command	Requires	Description
`sessions send --agent <id> --message <text>`	`openclaw.token`	Send message to agent via gateway
`sessions spawn --task <text>`	`openclaw.token`	Spawn a new agent session
`sessions list\|history\|status`	`openclaw.token`	Query sessions
`gateway trigger\|health\|agents`	`openclaw.token`	Gateway operations
`poll --rooms <r> --api-key <k> --handle <h>`	Ant Farm API key	Poll rooms for messages
`emit --to <url> --json <file>`	None (user specifies target)	POST event JSON to a URL
`hooks create --webhook-url <url>`	None (user specifies target)	Create webhook forwarder
`tmux run --cmd <command>`	Allowlisted commands only	Run command in tmux, capture receipt
`exec request\|resolve\|list`	Allowlisted commands only	Execution approval workflow
`cron add\|list\|remove\|run\|status`	`openclaw.token`	Scheduled task management

Configuration

Generated by ide-agent-kit init. All credential fields default to empty.

Field	Purpose	Default
`listen.host`	Webhook server bind address	`127.0.0.1`
`listen.port`	Webhook server port	`8787`
`tmux.allow`	Allowlisted shell commands	`[npm test, npm run build, pytest, git status, git diff]`
`openclaw.token`	Gateway auth (advanced commands)	empty
`github.webhook_secret`	Verify GitHub webhooks	empty

Data Access

Path	Access	Purpose
`ide-agent-receipts.jsonl`	append	Audit log of all agent actions
`ide-agent-queue.jsonl`	read/write	Event queue
`ide-agent-kit.json`	read	Runtime configuration (may contain secrets)
`memory/`	read/write	Local agent memory files

Source & Verification

npm: https://www.npmjs.com/package/ide-agent-kit
Source: https://github.com/ThinkOffApp/ide-agent-kit
Maintainer: petruspennanen (npm), ThinkOffApp (GitHub)
License: AGPL-3.0-only

The npm package contains no install scripts (preinstall/postinstall). All code is plain ESM JavaScript. Verify with npm pack --dry-run before installing.

ide-agent-kit

About