What problem does it solve?
Voice dictation transfers more context than typed prompts, but LLMs confidently fill every gap in a messy ramble. Agents then build on misreadings the user never sees until an hour of generated work is wasted. This Skill forces a structured echo brief before any action runs.
Core Features & Use Cases
- Echo Audit: Produces a five-section brief covering mission, locked decisions, open questions, flips and parked tangents, and the model's own inferences and guesses quarantined from the user's words.
- Capture Mode: Listens across multiple messages when the user invokes the skill before rambling, replying only with minimal acknowledgments until the user signals completion.
- Targeted Interview: Optionally asks one question at a time about open or guessed items, capped at five questions, then restates only the sections that changed.
- Use Case: A founder dictates a ten-minute stream of consciousness about building a daily Hacker News digest agent. The skill returns a scannable echo that surfaces the Slack-to-email reversal, quarantines the guessed source pool, and flags the undefined relevance ranking for correction before any code is written.
Quick Start
Install the thinking-out-loud skill and dictate a ramble about what you want to build, then review the echo brief and correct anything wrong before approving.
