API配额监控与手动切换
PassAudited by ClawScan on May 1, 2026.
Overview
The skill is coherent and matches its stated purpose, but users should know it can read OpenClaw session/config files, change the default model, and restart the Gateway when invoked.
This skill appears purpose-aligned, but treat model switching as an operational change: approve switching only when you are comfortable changing the default model and restarting the OpenClaw Gateway. Check the target model, know how to revert, and only enable the optional cron job if you want ongoing scheduled checks.
Findings (5)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Running the switch command can temporarily stop or restart the local OpenClaw Gateway.
The script kills and restarts the OpenClaw Gateway as part of model switching. This is disclosed in SKILL.md and fits the purpose, but it can interrupt service if run at the wrong time.
subprocess.run(["pkill", "-f", "openclaw-gateway"], capture_output=True)
Only approve --confirm or --model when you are ready for a Gateway restart, preferably outside important active sessions.
After switching, future OpenClaw requests may use a different model/provider than before.
The script modifies the OpenClaw configuration to change the default primary model. This is the advertised function, but it affects future agent/model routing.
config["agents"]["defaults"]["model"]["primary"] = model
Confirm the target model/provider before switching, and keep a backup or note of the previous model if you may want to revert.
The skill may inspect local OpenClaw session status/error data to decide whether a model switch is advisable.
The script reads OpenClaw session state to count recent errors and quota/rate-limit signals. It appears scoped and does not transmit the data, but session files can contain sensitive operational context.
SESSIONS_FILE = os.path.expanduser(os.environ.get("OPENCLAW_DIR", "~/.openclaw") + "/agents/main/sessions/sessions.json")Use it only on OpenClaw profiles where reading recent session error metadata is acceptable.
Installation metadata may not fully prepare users for the local tools and permissions needed.
The registry requirements do not declare local dependencies/config paths, while SKILL.md and the script require Python, OpenClaw configuration access, and Gateway restart commands. This is under-declared but the behavior is disclosed in the artifacts.
Required binaries (all must exist): none; Required config paths: none
Before use, verify Python 3, the OpenClaw CLI/Gateway, and access to ~/.openclaw/openclaw.json are available.
If you add the cron job, the monitor will keep running periodically and writing logs.
The documentation includes an optional cron-based recurring check. It only runs --check rather than switching models, and it is user-directed, but it is still a persistent scheduled action.
*/30 * * * * cd $HOME/.openclaw/skills/api-monitor && python3 api_monitor.py --check >> /var/log/api-monitor.log 2>&1
Only add the cron entry if you want recurring monitoring, and remove it from crontab when no longer needed.