Skip to main content
Before enabling Enforce mode, you want to see which guardrail templates match your agent’s risky actions. This recipe runs five scenarios — deploy, vault, database, merge, and shell — in Monitor mode. Source: guardrail-packs-smoke.ts · guardrail_packs_smoke.py

​
The five scenarios

#ToolActionResourceRisk
1deploy.releaseexecutedeployment_eventhigh
2vault.readreadsecrethigh
3db.updateupdatedatabase_recordhigh
4github.merge_prmergecode_repositoryhigh
5shell.rm_rfdeleteshell_commandcritical

​
Run the smoke test

await apie.withRun(
  { inputSummary: "Exercise starter guardrail packs in monitor mode" },
  async (run) => {
    for (const scenario of scenarios) {
      await apie.withTool(
        { runId: run.id, ...scenario },
        async () => ({ ok: true }),
      );
    }
  },
);

​
What you’ll see

Five agent.tool.called events and five agent.guardrail.evaluated events. Each evaluation shows the matched template and decision. In Monitor mode, all callbacks succeed regardless of decision.

​
Before running

Enable guardrail templates: 
npx apie guardrails enable prod-secrets
npx apie guardrails enable prod-deploys

​
Escalate to Enforce mode

  1. Review evaluation events in the dashboard
  2. Declare capabilities for all five tools
  3. Switch mode to "enforce"
  4. Re-run — blocked scenarios will throw before the callback executes
See Enforce guardrails.

​
Next steps

Monitor mode

How monitor mode works.

Enable guardrail templates

Turn on starter packs.