AgentSkillsCN

Logging

日志的存储位置与书写时机——错误日志、每日日志、系统日志

SKILL.md
--- frontmatter
name: Logging
description: Where logs live and when to write them — errors, daily, system
metadata:
  {
    "openclaw": {
      "emoji": "📋",
      "requires": {},
      "always": false
    }
  }

Logging

You have several log sinks. Use them so you (and Dmitry) can see what broke and when.

Where logs are

WhatWhereHow to read
Critical errorsdata/ERROR_LOG.mdread_file("data/ERROR_LOG.md") or tool log_error() to append
Display failuresdata/display_error.logread_file("data/display_error.log") — written by gotchi_ui.py on crash
Daily notesmemory/YYYY-MM-DD.mdwrite_daily_log(entry) to append; read via read_file
Code changesCHANGELOG.mdlog_change(description) to append
Service / systemjournalctlmanage_service(gotchi-bot, logs) or bash journalctl -u gotchi-bot -n 50
Conversationgotchi.db (messages)recall_messages(limit)

When to use what

  • •

    Something broke (display, service, disk, health):
    Call log_error("short description") so it’s in data/ERROR_LOG.md. Then you can say “last error was …” or read the file when asked.

  • •

    health_check() found problems:
    Call log_error("health_check: <what’s wrong>") so there’s a record.

  • •

    User says “что сломалось” / “what went wrong”:
    Read data/ERROR_LOG.md (and optionally data/display_error.log) and summarize.

  • •

    Daily / notable events:
    Use write_daily_log(entry).

  • •

    You changed code/config:
    Use log_change(description).

No direct system log “tool”

You don’t have a function that returns raw system logs; use manage_service(service, action=logs) or execute_bash("journalctl -u gotchi-bot -n 50") when you need recent service output.

Size limits (so we don't fill disk on Pi)

  • •ERROR_LOG.md — only last 300 lines kept; older lines are dropped when appending.
  • •display_error.log — only last 200 lines kept.
  • •Daily logs and CHANGELOG are not auto-trimmed; trim manually if needed.

Summary

  • •Errors → log_error(message) → data/ERROR_LOG.md
  • •Display crash → already in data/display_error.log (no tool needed)
  • •“When did something break?” → read data/ERROR_LOG.md