dag-agent

Configure, extend, and operate the Collar Agent CLI across platforms and providers.

1|Updated Jun 17, 2026
One-click install
npx skills add https://github.com/specdog/collar --skill dag-agent-specdog
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: dag-agent
Source: https://github.com/specdog/collar/tree/main/skills/dag-agent
Command: npx skills add https://github.com/specdog/collar --skill dag-agent-specdog

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Setting up and operating the Collar Agent framework involves many moving parts — providers, credentials, toolsets, gateways, MCP servers, cron jobs, and multi-agent spawning — and this Skill consolidates the commands, config keys, and troubleshooting steps needed to manage all of it. ## Core Features & Use Cases - CLI & Configuration Reference: Covers the full dag command surface — chat, setup, model selection, auth, profiles, sessions, cron, webhooks, and credential pools — plus every config.yaml section and provider env var. - Multi-Agent Spawning: Shows how to launch independent Collar instances via one-shot mode or tmux PTY sessions, coordinate parallel agents with worktree isolation, and choose between delegate_task and full process spawning. - Gateway, MCP & Durable Systems: Documents messaging platform setup (Telegram, Discord, Slack, and 15+ others), native MCP server configuration, webhook subscriptions, the curator skill lifecycle, and the kanban multi-agent work queue. - Use Case: You want an agent to monitor GitHub issues and post triage summaries to Telegram on a schedule — this Skill walks you through enabling the webhook platform, creating a subscription with a prompt template, and wiring delivery. ## Quick Start Ask the agent to run dag setup and configure a model provider, then verify the installation with dag doctor.

Frequently Asked Questions about dag-agent

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I set up Collar Agent with an LLM provider?▼

Run dag setup for the interactive wizard, or dag model to pick a provider directly. API keys go in ~/.dag/.env (e.g. OPENROUTER_API_KEY), while OAuth providers like Nous, OpenAI Codex, and Qwen are added with dag auth add.

How do I spawn multiple Collar agents in parallel?▼

Use tmux to launch independent interactive sessions, for example tmux new-session -d -s agent1 'dag -w', then send prompts with tmux send-keys. For quick subtasks, prefer the delegate_task tool, which has less overhead than a full process.

Does Collar support MCP servers?▼

Yes, Collar has a built-in MCP client. Add servers under mcp_servers in ~/.dag/config.yaml using either a stdio command or an HTTP url, install the mcp Python package, and tools are auto-discovered and registered as mcp_<server>_<tool> on startup.

Which messaging platforms does the Collar gateway support?▼

The gateway supports Telegram, Discord, Slack, WhatsApp, Signal, Email, SMS, Matrix, Mattermost, Home Assistant, DingTalk, Feishu, WeCom, BlueBubbles, Weixin, API Server, and Webhooks. Enable and configure platforms with dag gateway setup.

Why are my config changes not taking effect in Collar?▼

Many settings, including tool toggles and security.redact_secrets, are read once at startup. Change them via dag config set from a terminal, then start a fresh session with /reset or a new dag invocation.

When should I use delegate_task instead of spawning a dag process?▼

Use delegate_task for quick parallel subtasks that finish within the parent's turn, since it shares the process with isolated context. Spawn a full dag process for long autonomous missions needing full tool access, interactivity, or multi-hour runtimes.