T09 · Insecure Skill Coding Practices
Warning
- Location
scripts/get-diagnostic-info.sh:5- Finding
Unredacted OpenClaw Configuration and Log Disclosure
- Content
View full analysis
/dev/null; then openclaw logs -n $LOG_LINES 2>/dev/null || echo "❌ Unable to retrieve logs" else echo "❌ The openclaw command is unavailable" fi } case "${1:-all}" in config) get_config ;; logs) get_logs ;; status) get_status ;; all|*) get_config echo "" get_status echo "" get_logs ;; esac ``` The default diagnostic workflow is also explicitly recommended in `SKILL.md` lines 15–25 and `references/common-issues.md` lines 8–12. ### Technical Analysis The script outputs the entire `~/.openclaw/openclaw.json` file through `cat` and retrieves the most recent 100 log lines without applying any redaction or field selection. OpenClaw configuration and logs may contain authentication tokens, API keys, bot credentials, webhook URLs, account identifiers, message content, internal addresses, or other sensitive operational data. Because the default argument is `all`, invoking the script without an explicit mode exposes both the complete configuration and recent logs. The project does not contain an explicit mechanism that transmits this output to an attacker. However, when the script is run as part of an AI-assisted diagnostic workflow, its output can enter the Agent context, conversation transcript, telemetry, or a user-visible response. This v ...[truncated 1333 chars]- Remediation
View remediation