do-bitter-pill

Run the QuickCheck workflow on the file to identify independently actionable rules, apply the six tests for defaults, contradictions, redundancy, one-off fixes, vagueness, and reasoning choreography, then receive a verdict listing rules to cut, sharpen, or keep.

Over-prompting is adding rules that restate behavior the model already enforces, duplicate other instructions, preserve one-off workarounds, or script reasoning steps where an outcome or constraint would suffice. The core test is whether a more capable model would make the rule unnecessary.

No, the audit is read-only by default. It produces a classification report with recommended actions, and files are only changed after you explicitly request edits.

Use the full Audit when reviewing an entire instruction hierarchy across a repository, including cross-source conflicts, loading evidence, and context-value estimates. Use QuickCheck for a single file or an inline rule block.

do-bitter-pill audits instruction sets for over-prompting and rule hygiene, while do-red-team attacks the logic of an idea. Use do-bitter-pill to trim setup files, not to critique arguments.