Backup of conversations to Obsidian

Automatic conversation backup system for Obsidian with incremental snapshots, hourly breakdowns, and formatted chat-style markdown. Use when setting up conversation archival, preventing data loss from /new resets, or organizing chat history in Obsidian vault with proper formatting (colored callouts, timestamps, multi-paragraph support).

This package appears to implement the advertised backup behavior, but it has sloppy configuration handling and will try to read local config files for a Telegram token even though no credentials are declared. Before installing: (1) open and inspect install.sh and the scripts (monitor_and_save.sh, save_full_snapshot.sh) yourself — do not run blindly; (2) verify or change default paths (they default to /root/…); run/install as a non-root user if possible; (3) if you do not want Telegram warnings, remove or disable the curl blocks or ensure BOT_TOKEN/CHAT_ID are empty; (4) confirm where CHAT_ID should be set (scripts do not read the included config.example) and either add explicit config parsing or set environment variables yourself; (5) test the scripts on a small sample session file to confirm they only write local snapshot files and do not transmit conversation contents; (6) consider moving tracking files into a directory you control and audit permissions. If you want, I can point out the exact lines to change to remove Telegram calls and hardcoded root paths.

Type: OpenClaw Skill Name: obsidian-conversation-backup Version: 1.0.0 The skill is classified as suspicious due to the `scripts/monitor_and_save.sh` script accessing the sensitive `/root/.clawdbot/clawdbot.json` file to retrieve a Telegram bot token and then performing network egress via `curl` to send messages to the Telegram API. While this is documented as a 'token warning' feature and the messages sent are predefined, the combination of reading a sensitive configuration file and initiating external network communication represents a high-risk capability that could be misused, even if not explicitly malicious in its current implementation.