Security audit
Security checks for vulnerabilities and agentic risk
This is a Markdown-only guardrail skill for cron and background-worker reliability, with no hidden code, dependency installation, exfiltration, or persistence behavior found.
Review the skill as reliability guidance for unattended jobs, not as a complete implementation. The referenced files under references/ are not present in this artifact, so users should not assume extra hardening material is installed. Use exec-enabled workflows only in repos you intend the agent to inspect or modify.
This markdown includes substantial natural-language content in Chinese for core capability, scenarios, differentiation, and trigger keywords, while other sections are in English. The file does not indicate that the skill is region-specific or give users an option to choose language, which can violate a language/locale policy requiring user opt-in.
The manifest description and scope sections frame the skill as POSIX-focused hardening guidance for OpenClaw cron/background workers, especially bash/sh usage. Later documentation expands applicability to Windows and arbitrary AI agent platforms, which broadens the advertised behavior beyond the manifest's stated scope.
