What problem does it solve? Converting an existing Custom GPT into a portable Agent Skill is risky: instructions, knowledge files, actions, and hidden platform behavior do not map one-to-one, and naive copying causes silent semantic loss. This Skill produces a structured conversion plan with capability mapping, semantic-loss tracking, and a migration backlog before any authoring begins. ## Core Features & Use Cases - Capability Mapping: Inventories GPT instructions, knowledge files, actions, starters, and prompt chains, then maps each behavior to a skill construct (trigger, procedure, reference, script, adapter, or drop). - Semantic-Loss Analysis: Identifies platform-bound capabilities like connectors, OAuth actions, conversation memory, and hidden Builder behavior, pairing each loss with a mitigation or acceptance test. - Foundry Handoff: Emits a machine-readable conversion dossier and exactly three eval cases so okhp3-skill-foundry can author, benchmark, and iterate on the final SKILL.md. - Use Case: You have a GPT that reviews architecture decision records using a rubric, a glossary PDF, and a six-stage prompt chain. Use this Skill to produce a conversion verdict, capability map, and migration backlog before rebuilding it as a portable skill. ## Quick Start Plan the conversion of my Custom GPT into an Agent Skill based on its instructions, knowledge files, and actions.
