Writing a regex validation library is exactly the kind of precise coding task AI handles well.

The task structure is fully deterministic: produce a JS module with specific patterns and a test suite. There is no judgment variation between instances — the same inputs should always produce the same outputs.

Success criteria are crisp and verifiable: patterns must match valid inputs and reject invalid ones, and tests must pass. The only mild ambiguity is in 'corporate domain filtering,' but that is a well-understood concept with standard heuristics (e.g., blocking gmail.com, yahoo.com).

All required knowledge — AWS ARN format specs, Stripe key prefixes, E.164 phone formats, RFC 5321 email rules — is in the agent's training data. No external APIs, credentials, or file access are needed.

Output is a code artifact that a developer reviews before use. False positives or missed edge cases are caught during code review or QA, and the module is trivially correctable. No irreversible consequences.

Regex authoring is a technical skill with objective correctness criteria. The only subjective element — which free email domains to block — has well-established community lists the agent can reference.