name: openclaw-usage-ui description: Read and format OpenClaw Control UI “使用情况” metrics using the same internal UI-style usage source. Use when the user asks for today’s model usage, hourly/recent usage, per-model input/output/cacheRead totals, or wants OpenClaw usage output rendered in a fixed template. Prefer this skill when the user explicitly wants numbers aligned with the UI “使用情况” view rather than rough transcript or trajectory estimates.

OpenClaw Usage UI

Use the bundled script to read usage from assistant message usage fields in session transcripts. This matches the UI-style “使用情况” view much better than trajectory model.completed aggregation.

Quick Start

Run:

python3 /Users/mac/.openclaw/workspace/skills/openclaw-usage-ui/scripts/render_usage_ui.py

The script outputs the user-facing template directly.

Rules

• Treat transcript message.usage as the source of truth for this skill.
• Count only assistant messages with usage payloads.
• Use input, output, and cacheRead fields.
• Do not invent “模型三” or later blocks when there is no data.
• Keep the output template exactly in the user-requested style unless they ask for a different format.
• If the user asks for a different time window later, update the script or invoke a variant rather than reverting to trajectory estimates.

Output Template

Use this structure:

模型一：<provider/model>

今日总输入量（M）：<value>

今日总输出量（M）：<value>

今日总缓存命中量（M）：<value>

最近一小时：输入量（M）<value>、输出量（M）<value>、缓存命中量（M）<value>

Repeat only for models that actually have data.

Notes

• The current script uses Asia/Shanghai time.
• The current script scans /Users/mac/.openclaw/agents/main/sessions/*.jsonl and ignores *.trajectory.jsonl.
• Values are rendered in M with 6 decimal places.