Measure whether the tool registry runtime works

Translate the claim into assertions or a tiny eval harness. The metric to protect is: allowlist reject count; schema validation failures; successful typed results. Always record the denominator (how many cases) beside any rate. A percentage without a denominator is marketing, not measurement.

const parsed = OutputSchema.safeParse(raw);
if (!parsed.success) {
  throw new ContractError("invalid_output", parsed.error.issues.map(i => ({
    path: i.path.join("."), code: i.code
  })));
}
return parsed.data;

Expected evidence: duplicate names, stale schemas, timeouts, oversized results, and unsafe side effects. A passing assertion proves only the behavior it names; broader usefulness still needs the chapter’s full limits.

Be explicit: beating the baseline (inline if/else tool routing without schemas) on this fixture does not prove behavior under string-eval calculator, or lookup that returns entire DB on missing id. Label observations separately from conclusions so the next page inherits honest evidence about the tool registry runtime.

Compute allowlist reject count; schema validation failures; successful typed results with the denominator written beside the rate every time. For this chapter, the evaluation set is intentionally tiny; that is allowed only if you say so in the evidence. Compare against inline if/else tool routing without schemas before celebrating.

Add one negative case aimed at string-eval calculator, or lookup that returns entire DB on missing id. A suite with only happy cases cannot protect the tool registry runtime when the characteristic failure appears in review.

If a check is expensive or flaky, shrink it until it is deterministic on registry with calc and kb.lookup. Flaky green builds teach the team to ignore gates. Record what this page does not prove so security-ops and mastery-ship inherit honest limits.

At the evaluation stage for tools-in-code, the job is narrower than finishing a product demo. You are creating one progressive evidence piece about registry with calc and kb.lookup that later pages inherit without redefining success. Keep that fixture small enough to inspect by hand, keep outputs copy-pasteable as text, and refuse to narrate this baseline as if it were a production SLA: inline if/else tool routing without schemas.

For this page specifically, success looks like metrics with explicit denominators and a negative case while still centering the user decision to route model tool requests through a registry that validates names and args. If you cannot point to a file, command, or assertion that proves that for the tool registry runtime, stay on this page instead of advancing.

Glossary: tool · Glossary: structured output · Cheatsheet: production ops signals

Previous · Next

For tools-in-code, write a transfer example that differs in one constraint from the chapter scenario. Keep the quality bar fixed. Explain which check still applies.