What problem does it solve? When you need a second opinion or want to continue work in ChatGPT, Ollama, Gemini, or another LLM, that model has zero context about your repo, session, or task. This Skill packages the current work into one self-contained master prompt so a cold, tool-less model can act on it immediately. ## Core Features & Use Cases - Structured 9-section master prompt: Builds a handoff covering role, mission, context, current state, verbatim artifacts, constraints, task, output format, and open questions. - Dual target modes: Emits a full self-contained prompt for cold targets (browser ChatGPT, plain ollama run) or a lean orchestration brief for agentic targets (Codex, opencode, Aider) that can read the repo themselves. - Local, git-ignored archive: Saves each handoff to .claude/workspace/reports/transfers/{tool}/ as a dated Markdown file that is never committed. - Use Case: You hit a stubborn bug and want ChatGPT's take. Invoke the Skill, and it reads the real files, embeds the exact code and error output, and prints a copy-paste block plus a saved .md you can pipe into ollama run. ## Quick Start Ask the assistant to transfer the current task to ChatGPT or Ollama, and it will gather the relevant context and produce a ready-to-paste master prompt saved under the transfers folder.
