Agent Audit

This skill appears to be what it says: a read-only audit that reads your OpenClaw config and (when available) cron/session history to estimate token usage and recommend model tier changes. Before running: (1) review scripts/audit.py yourself (it runs locally and will read files under ~/.openclaw and related config locations); (2) run with --dry-run and --output to inspect results without making changes (the tool claims it never writes config automatically); (3) be aware that the tool estimates costs using the included pricing table — it does not access provider billing APIs or require API keys, so its cost estimates are approximate unless you supply cron/session histories that include accurate token counts; (4) if you plan to grant it access to an OpenClaw cron API or other telemetry, understand what those endpoints expose (session contents, logs, or tokens may contain sensitive data). If you want absolute assurance, run the script in a safe environment or audit the code line-by-line; otherwise the skill is coherent and proportionate for its purpose.