writing-skills

Test-driven skill authoring applies RED-GREEN-REFACTOR cycles to process documentation, using baseline pressure scenarios to capture and harden against agent rationalizations before finalizing instructions. It ensures agent behavior matches documented rules.

Test-driven skill authoring is a method that treats documentation like code. It requires capturing baseline failures without the skill, writing minimal instructions to pass, then identifying new rationalizations and adding explicit negations to re-verify compliance.

Verify agent compliance under pressure by running explicit pressure scenarios that test time, sunk cost, authority, and exhaustion constraints. This loophole hardening process identifies rationalizations and tests that agents still follow documented rules.

Test-driven skill authoring requires no external dependencies or frameworks. It relies entirely on running baseline pressure scenarios against your agent, observing rationalizations, and updating minimal instructions to harden process documentation compliance.

The best way to prevent loopholes is loophole hardening: after passing initial tests, identify new rationalizations agents use under pressure, add explicit negations or counters to the instructions, and re-test until the rule remains intact.

Avoid TDD for process documentation when rules are highly subjective or lack clear pass/fail pressure scenarios. The method requires explicit baseline failures and observable rationalizations to successfully harden agent compliance.