Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed robotaxi news-briefing workflow with public-source fetching and scoped short-term report storage for delta comparison.
Install only if you want an agent to fetch public robotaxi-sector sources, write brief report files under your OpenClaw memory directory, and clean up old matching robotaxi report files. Review the shell examples if your deployment limits shell execution or requires user approval before local writes.
Lines L143-L147 state that for/while loops, variable assignment, pipelines, and multi-line shell scripts are 'absolutely forbidden' for shell safety. However, the earlier mandatory RSS scan section uses a multi-line bash script with a for loop, variable assignment, a curl pipeline, and inline Python, directly contradicting the stated safety guidance.
The manifest description broadly says the skill tracks robotaxi intelligence and generates briefings, but it does not specify explicit invocation phrases, scope limits, or exclusion conditions. In a manifest file, this kind of open-ended capability description can create ambiguous activation boundaries for when the skill should or should not be invoked.
The skill embeds substantial shell scripting and network-fetch logic, including curl pipelines and Python execution, which exceeds the minimum needed for a news briefing workflow and increases the attack surface. Even if intended for data collection, encouraging arbitrary shell-capable execution in a skill creates risk of command misuse, environment interaction, and unsafe extension by downstream agents.
The skill instructs persistent local writes and deletion of files under a user directory without a clear user-facing warning or consent boundary. This creates risk of unintended state changes, report retention of fetched content, and destructive cleanup behavior that could remove data unexpectedly if path assumptions are wrong or reused unsafely.
The skill explicitly persists reports to a shared memory-like path and retains them for 24 hours, creating session persistence that can leak prior-run context, fetched content, or derived intelligence across executions. In multi-tenant or shared-agent settings, this can expose sensitive operational history and creates a foothold for poisoning future delta comparisons using tampered prior files.
Before outputting the report to chat, save it with write:
write ~/.openclaw/workspace-group/memory/robotaxi_report_YYYYMMDD_HHMM.md
Then delete files older than 24h:
The instruction mandates 'All output in Chinese' as a fixed language policy. This is a natural-language locale restriction that does not offer the user a choice or explain a justified region-specific requirement.
No suspicious patterns detected.