界面已本地化;技能介绍与说明文档按所选语言由 AI 自动翻译,代码、命令、路径和专业术语保留原文,以原作者文档为准。
探索技能
效率与协作 · SKILL.md

live-watch

live-watch

【效率与协作】Polls the Agent Monitor /api/stats endpoint several times over a short window and reports the live deltas in active_sessions, active_agents, events_today, and ws_connections so you can see activity moving in real time. Use when watching the dashboard for live changes rather than a one-time snapshot.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准

技能说明

GitHub

name: live-watch description: > Polls the Agent Monitor /api/stats endpoint several times over a short window and reports the live deltas in active_sessions, active_agents, events_today, and ws_connections so you can see activity moving in real time. Use when watching the dashboard for live changes rather than a one-time snapshot.

Live Watch

Watch the dashboard's live counters change over a short window by polling /api/stats a few times and reporting the deltas.

Input

The user provides: $ARGUMENTS

Interpreted as the watch shape: number of polls and/or interval (e.g. 5x3s = 5 samples 3 seconds apart). Defaults when empty: 5 samples, ~3 seconds apart (a ~15-second window). A bare number means that many samples at the default interval; a bare duration means the default sample count at that interval.

Data Sources

EndpointReturns
GET /api/stats (polled){ total_sessions, active_sessions, active_agents, total_agents, total_events, events_today, ws_connections, agents_by_status, sessions_by_status }

Method

Poll GET /api/stats once per interval for the configured number of samples, recording the timestamp and the four watched counters each time. Pace the polls with a short wait between requests; keep the total window short (seconds, not minutes) so it stays interactive.

If the very first poll fails to connect, the dashboard is down — stop and tell the user to start it with npm start (or npm run dev) from the repo root, then retry.

Report Sections

1. Watch Window

State the sample count, interval, and total elapsed window.

2. Sample Timeline

A Markdown table — one row per poll — with columns: #, time, active_sessions, active_agents, events_today, ws_connections.

3. Deltas

For each of the four watched counters, report the net change from the first to the last sample using ▲ (increase), ▼ (decrease), or = (no change). Note any mid-window spikes or dips visible in the timeline.

4. Verdict

One line: is the dashboard actively receiving traffic (counters moving) or idle (flat) over the window?

Output

  • Compact Markdown. The timeline table is the centerpiece.
  • Cite real values from each poll — never interpolate or invent samples.
  • Deltas use ▲/▼/= with the signed numeric change, e.g. events_today: ▲ +7.
  • Keep it scannable in a terminal — no padding beyond the table.
我的收藏