By name: log-analyzer description: "Analyze picoclaw application logs. Use proactively after any code change, deployment, or server restart — do not wait for the user to ask. Also use when debugging unexpected behavior, after API calls fail, or when the user mentions errors or system health."
log-analyzer Skill
Analyze picoclaw application logs on demand using the logs tool.
Token-Efficient Analysis Pattern
Always start narrow, then widen only if needed:
- Overview —
logswithlevel=WARN,limit=20to get a quick summary of issues - Drill down — If errors found, use
componentorqueryto isolate - Full context — Only use
level=INFOorlevel=DEBUGwhen the user explicitly asks or when WARN/ERROR alone is insufficient
Never fetch DEBUG-level logs without a specific reason — most are routine noise.
logs Tool Parameters
| Parameter | Default | Use when |
|---|---|---|
level |
WARN | User says "errors" → ERROR, "everything" → DEBUG |
component |
(all) | User mentions a specific channel: telegram, discord, slack, wecom, agent |
limit |
50 | Start with 20 for overview, increase only if needed |
query |
(none) | User mentions a keyword: "timeout", "webhook", "auth" |
Analysis Checklist
When reporting findings:
- Error count & types — Group by component and error pattern
- Timestamps — Note if errors are clustered (burst) or spread out
- Recurring patterns — Same error repeating = systematic issue
- Actionable items — Suggest concrete fixes, not just descriptions
User Intent Mapping
| User says | Parameters |
|---|---|
| "Check the logs" / "Any errors?" | level=WARN, limit=20 |
| "What's wrong with Telegram?" | level=WARN, component=telegram |
| "Show me everything" | level=DEBUG, limit=50 |
| "Find timeout errors" | level=ERROR, query=timeout |
| "System health check" | level=WARN, limit=30 then summarize |
Report Format
分析後は必ず以下の形式で出力する:
ログ調査結果
- 調査範囲: WARN以上 / 直近N件
- エラー: X件 警告: Y件
- 影響コンポーネント: [一覧 または「なし」]
検出された問題 (なければ省略)
- [component]
エラーパターン— N回発生、最初: T / 最後: T → 推定原因 / 対応案
判定
- 異常なし — 対応不要
- 要対応 — 上記の問題を確認してください