openclaw-audit-watchdog
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONPERSISTENCEDATA_EXFILTRATIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes system commands and internal scripts using
openclawCLI,noderuntimes, andbashto perform audits, manage cron jobs, and deliver reports. These actions are limited to the skill's administrative tasks. - [PERSISTENCE]: The skill uses the
openclaw cronutility to create or update an unattended recurring job named "Daily security audit (Prompt Security)". This persistence is a declared core feature for automated monitoring. - [EXTERNAL_DOWNLOADS]: The skill fetches artifact verification data from the vendor's official GitHub repository to ensure archive integrity. It also supports optional remote updates via
git pullfrom the vendor repository when configured. - [DATA_EXFILTRATION]: Security reports are delivered to external targets (Telegram, Slack, or Email) as configured by the user. The skill implements header sanitization (stripping newlines) and SMTP-specific safety measures to prevent delivery abuse.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted audit data from external sources (other skills) to generate summaries.
- Ingestion points: Findings are read from JSON files by
scripts/render_report.mjs. - Boundary markers: The skill uses structured JSON parsing rather than raw text interpolation.
- Capability inventory: Network delivery via message tools and SMTP/sendmail.
- Sanitization: Implements newline stripping in mail headers and SMTP dot-stuffing.
Audit Metadata