Verification Before Completion

Quick Review

Excellent skill that addresses a critical failure mode in AI agent workflows. The description clearly conveys the core principle (evidence before claims), and the SKILL.md provides comprehensive, actionable guidance including the 5-step Gate Function, specific examples of proper/improper verification patterns, common rationalizations to avoid, and red flags. Structure is logical with clear sections and tables. This is highly novel because it tackles behavioral/cognitive patterns where LLMs commonly fail (premature success claims, insufficient verification), and provides the specific mental framework and checklists needed to prevent these failures. The reference to 24 failure memories adds concrete motivation. Minor room for improvement: could explicitly list the exact commands for common scenarios (test runners, linters, build tools by language), though the current approach is appropriately general. This skill meaningfully reduces the token cost and error rate of ensuring proper verification discipline.