What problem does it solve? When a user request is ambiguous, a test fails unexpectedly, or CLAUDE.md and a skill seem to disagree, AI coding assistants often guess silently or overreach in scope. This Skill supplies a full decision table with per-row rationale so every judgment call under uncertainty follows a documented, defensible default. ## Core Features & Use Cases - Thirteen decision rules with rationale: Covers bug-fix scope, missing test infrastructure, suspicious failing tests, performance and cleanup requests, CLAUDE.md vs skill conflicts, approval ambiguity, confidence scoring, and more. - Explicit overrides and anti-patterns: Each rule states when the default does not apply, plus a list of behaviors to avoid such as silently picking between interpretations or rounding up confidence scores. - Use Case: A user says "make it faster" with no details. Instead of speculatively adding caching, the assistant follows rule 4: profile first, identify the bottleneck, then propose a targeted change with measured impact. ## Quick Start Ask the assistant to apply the decision-rules skill to decide how to handle an ambiguous request like a failing test that looks wrong.
