trustlog-guard
anouartrust 作
Coding Agents & IDEs概要
Financial governance for OpenClaw agents.
説明
このスキルでできること
Reads OpenClaw agent session logs to track API spending by model and session. Enforces daily and monthly budgets, detects anomalies like runaway loops and model escalations, and produces formatted cost reports. All processing is local with no external connections.
Unlike cloud-based cost monitoring tools, it reads spend data directly from local session files with no external accounts, API keys, or data transmission required.
使いどころ
- Reviewing total API spend after a long agent session
- Setting a daily budget limit to prevent unexpected charges
- Spotting an agent stuck in a message loop before costs escalate
- Identifying which AI model is consuming the most budget
- Finding which sessions drove the highest spending this month
ワークフロー例
User asks: /trustlog
- 1
Scans ~/.openclaw/agents/{agent}/sessions/*.jsonl for cost records
- 2
Groups message costs by today, this week, and this month
- 3
Ranks top sessions by total spend and message count
- 4
Runs anomaly checks for burn rate spikes, rapid-fire loops, and model escalations
- 5
Appends optimization tips based on model usage patterns and cache efficiency
Full financial report with spending overview, cost breakdown by model, top sessions, anomaly scan results, and one actionable optimization tip
インストール
clawhub install anouartrust/trustlog-guardCLIコマンドを実行するか、Claude Code、Cursor、Codexなどお使いのコーディングエージェントにプロンプトを貼り付けてください。