Back to directory

blurt

A screen-recording and voice tool that turns a spoken walkthrough into separate bug tickets, ideas and to-dos for a coding agent, picking the right frame for each item and writing up expected versus actual behavior with likely code pointers.

blurt

blurt records a user's screen while they talk through what they see, then hands the recording to a coding agent, which splits a rambling narration into separate items even when the speaker jumps between topics. For each item it picks the right video frame, marks the exact spot the mouse was pointing at, and writes it up with actual versus expected behaviour, reproduction steps, an owner, and the code that is likely responsible.

Items can be triaged one at a time from a keyboard-driven review page, then filed to Feishu or Lark, GitHub, Linear or a Markdown file, or handed straight to the agent for fixing. It installs as an agent skill for Claude Code, Codex and other skill-running agents, alongside a macOS menu-bar app for recording, and does not require a code repository to record a session, so a designer, product manager or client can record one and send the video to a developer's agent.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.