From agents-observe
Manages Agents Observe server: start/stop/restart, view dashboard, check stats, tail logs (Docker, CLI, MCP), and diagnose issues.
How this skill is triggered — by the user, by Claude, or both
Slash command
/agents-observe:observeThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Agents Observe dashboard and server management.
Agents Observe dashboard and server management.
/observe view — Open the current session in the dashboard/observe stats — Open the current session's stats modal in the dashboard/observe — Open the dashboard URL/observe status — Show server health and config details/observe start — Start the server/observe stop — Stop the server/observe restart — Restart the server/observe logs-server — Show recent Docker container logs/observe logs-cli — Tail the local cli.log file/observe logs-mcp — Tail the local mcp.log file/observe debug — Diagnose server issues (health, docker logs, mcp.log, cli.log)All commands are invoked via the wrapper at scripts/cli.sh, which resolves the path to observe_cli.mjs relative to this skill's install location. Do not reach outside the skill dir with ../ paths — the agentskills spec disallows it and it breaks portability across agents.
The subcommand is in $ARGUMENTS. If empty, default to showing the dashboard URL.
Opens the current session in the dashboard.
scripts/cli.sh health
Dashboard: URL (e.g. http://localhost:4981). If exit code 1, the server isn't running — tell the user to run /observe start and stop here.<dashboard>/#/_/${CLAUDE_SESSION_ID} (the _ is the project placeholder; the dashboard resolves the real project from the session id).open <url> on macOS, xdg-open <url> on Linux, start <url> on Windows. Pick based on the Platform: line in your environment context.Opens the current session's stats modal in the dashboard, using a deep-link URL.
scripts/cli.sh health
Dashboard: URL (e.g. http://localhost:4981). If exit code 1, the server isn't running — tell the user to run /observe start and stop here.<dashboard>/#/_/${CLAUDE_SESSION_ID}:session.statsopen <url> on macOS, xdg-open <url> on Linux, start <url> on Windows. Pick based on the Platform: line in your environment context.scripts/cli.sh health
/observe start or /observe status.scripts/cli.sh health
/observe restart./observe start.scripts/cli.sh start
scripts/cli.sh stop
scripts/cli.sh restart
scripts/cli.sh logs-server -n 50
-f (follow) — it would hang.scripts/cli.sh logs-cli -n 50
scripts/cli.sh logs-mcp -n 50
Run these checks in sequence. Read each output before running the next — use what you learn to diagnose the issue.
Server health:
scripts/cli.sh health
Docker container logs (last 20 lines):
scripts/cli.sh logs-server -n 20
MCP log (last 20 lines):
scripts/cli.sh logs-mcp -n 20
CLI log (last 20 lines):
scripts/cli.sh logs-cli -n 20
Analyze the results and tell the user:
npx claudepluginhub simple10/agents-observe --plugin agents-observeDebug container agent issues including failures, authentication problems, and sandbox runtime behavior. Covers logs, environment variables, mounts, and common fixes.
Starts a local dashboard server (Deno) for managing projects, viewing workflow graphs, agent streams, and documents. Useful for monitoring and managing multi-project workflows via browser.
Sets up evaluators, online monitoring, CI/CD quality gates, CloudWatch/X-Ray observability, and cost optimization for AgentCore agents to measure and improve quality and performance.