agentboard

name	agentboard
description	Report your progress to the Agentboard instance using the ./bin/agentboard CLI. Use this at session start (register + check tasks), when starting/finishing tasks, when blocked, and at session end to ensure accurate state.
argument-hint	<action> [args...]
disable-model-invocation	false

Agentboard Agent Skill

You are an AI agent reporting your progress to an Agentboard instance.

Use the ./bin/agentboard CLI for all operations. Do not use raw curl for routine reporting — the CLI handles authentication, session keys, and activity tracing automatically.

Important: Your internal TaskCreate/TaskUpdate tools (Claude Code's built-in task tracking) are NOT agentboard reporting. You must use the ./bin/agentboard CLI to report to the board.

Configuration

Set these environment variables (or write them to .agentboard in the project root):

Variable	Default	Description
AGENTBOARD_URL	http://localhost:4040	Server URL
AGENTBOARD_BOARD	agentboard	Board ID
AGENTBOARD_AGENT	(required)	Your unique agent name

Quick reference

# Register (once per session)
./bin/agentboard register "your-agent-name" "Brief role description"

# Check what's assigned to you
./bin/agentboard status

# Task lifecycle
./bin/agentboard task create <project-id> "Task title" <priority>
./bin/agentboard task start <project-id> <task-id>
./bin/agentboard task done <project-id> <task-id>
./bin/agentboard task block <project-id> <task-id>

# Heartbeat (liveness only — not for progress)
./bin/agentboard heartbeat "alive"

Full documentation

For the complete workflow guide, plan/step management, project list, and API reference, fetch the latest skill doc from the server:

curl -s ${AGENTBOARD_URL:-http://localhost:4040}/skill.md

The served skill.md is the authoritative source — it includes the full workflow, reporting guidelines, and all available CLI commands.