What problem does it solve? When an AI agent or skill produces output, there is no built-in way to verify the work actually matches the stated goal before the user sees it. This skill acts as a critical second reviewer that catches placeholder text, missing fields, logical contradictions, and convention violations before they reach the user. ## Core Features & Use Cases - Dual-mode operation: Runs as a subagent spawned by other skills via a structured check: block, or directly by the user with a natural-language description of what was done. - Four-step validation protocol: Restates the goal, checks each artifact for existence, completeness, correctness, and convention compliance, then evaluates every provided check explicitly. - Structured verdict report: Produces a table of PASS/WARN/FAIL results per check plus an overall verdict with specific actionable issues, including file paths and line references. - Use Case: After a skill generates a new SKILL.md file, spawn impl-validator with the goal and artifact list to confirm the frontmatter contains name and description, no placeholder text remains, and conventions are followed. ## Quick Start Ask the agent to validate the implementation it just produced by describing the goal and the files it created.
