Security audit
Security checks for vulnerabilities and agentic risk
The skill is purpose-aligned, but it automatically records cross-agent usage history in persistent OpenClaw memory without a user opt-in.
Install only if you are comfortable with this skill reading token and cost metadata for every registered OpenClaw agent and keeping an append-only usage history in OpenClaw workspace memory. There is no evidence of network access, credential use, executable code, or destructive behavior, but the automatic history file may reveal operational patterns to later tools or users with workspace access.
The skill is designed to append a structured history record containing cross-agent session usage data to a persistent workspace file on every run, even when the user did not request retention. This creates silent data accumulation and can expose operational metadata, model usage, spend patterns, and agent identifiers to later readers or other tools with workspace access.
The skill explicitly instructs the agent to always persist per-run summaries for all discovered agents with session data, regardless of whether the user asked for logging. Because the scan scope is all registered agents, this can aggregate and retain broader-than-expected metadata across multiple agents, increasing privacy and cross-tenant information exposure risks.
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
Use `null` for any numeric field where all entries in the group had the field absent. Use `[]` for `flags` when no flags apply. Agents with no session data are omitted from the record entirely.
## Step 8: Optional log write
**Only if the user explicitly requests logging**, append a timestamped summary to:
No suspicious patterns detected.