What problem does it solve? Rough, mixed-language, dictated, or repetitive coding requests often lose critical technical details when passed to a coding agent, causing the agent to misunderstand the task or start in the wrong place. This Skill reconstructs the user's real intent, preserves exact technical literals, and packages the result as a repo-aware handoff prompt. ## Core Features & Use Cases - Intent Reconstruction: Cleans filler, duplicates, and false starts from noisy input while preserving exact file paths, CLI commands, config keys, and code identifiers verbatim. - Task Mode Classification: Routes each request to the right mode (implementation, bug-investigation, review-read-only, refactor, planning-spec, architecture-analysis, docs-and-messaging, or tooling-prompting) based on readiness and risk rules. - Repo-Aware Context Injection: Selects only the repository facts, docs, and code areas that materially change the downstream agent's first actions, avoiding generic repo dumps. - Use Case: Paste a rambling voice-transcribed bug report mentioning opencode.json and provider.gonkagate, and receive a structured bug-investigation handoff prompt with exact literals, likely code areas, and explicit open questions. ## Quick Start Ask the agent to turn your rough coding notes into a context-rich handoff prompt for the opencode-setup repository.
