- name
- openclaw-cost-guard
- description
- This skill should be used when the user asks to reduce OpenClaw token spend, audit model and cron cost risk, prevent denial-of-wallet incidents, add budget guardrails, review expensive agent defaults, or tighten AI-agent cost governance before wider rollout.
OpenClaw Cost Guard
Use this skill to identify obvious cost leaks before they turn into a denial-of-wallet problem.
Goal
Move from “the agents work” to “the agents work within an intentional budget.”
Workflow
- Identify the target OpenClaw config or workspace.
- Run the bundled cost-guard script.
- Review expensive defaults, missing budgets, and high-risk patterns.
- If the result is
FAIL, stop scaling usage until cost controls are tightened. - Apply the narrowest guardrails that reduce spend without breaking the workflow.
Command
node {baseDir}/scripts/cost-guard.mjs --config ~/.openclaw/openclaw.jsonThe --config flag is optional. If omitted, the script checks the default OpenClaw config path.
What the script checks
- whether explicit budget signals exist
- whether default models look expensive for always-on usage
- whether browser/interactive tooling appears enabled without cost discipline
- whether token limits appear excessively large
- whether the config contains multiple high-cost patterns at once
- whether the setup needs a governance recommendation before scaling
Output format
The script returns JSON with:
scoreverdictsummaryfindingsrecommendationsguardrailsevidence
Verdicts
PASS— no major lightweight cost-governance gap foundWARN— spend risk exists and should be reviewedFAIL— denial-of-wallet risk is materially elevated
Important limits
- This is a lightweight static review, not a bill-reconciliation system.
- A low-risk config can still become expensive through user behavior or external automations.
- Always verify against real provider invoices and usage telemetry.
References
{baseDir}/references/cost-playbook.md