ryanlchan/hmm

Overview

This skill sends the most recent LLM response to a browser-based feedback and commenting UI for review. It opens a local editor with inline track changes, comment threads, and a structured export for edits and annotations. Use it when you want collaborative review, inline revisions, or an auditable summary of feedback.

How this skill works

The skill writes the last LLM response to a temporary preload file and launches a local web server that loads that file into an editable interface. The UI supports inline track changes, selectable text comments, and a Summarize action that exports all edits and comments as structured markdown. Once the server starts, the interface is available at http://localhost:8723 for desktop and mobile layouts.

When to use it

• You want reviewers to annotate and suggest edits directly on an LLM response.
• You need inline track changes and comment threads for collaborative review.
• You want a structured export of comments and edits for handoff or documentation.
• You prefer a lightweight local interface without cloud upload of content.
• You need a mobile-friendly review workflow with bottom-sheet comments.

Best practices

• Keep the last LLM response concise before sending so reviewers focus on the final content.
• Run the skill from the project root so the serve script can locate its assets.
• Use the Summarize export to capture all comments and edits as markdown for records or next-step work.
• Encourage reviewers to use inline comments for context and the track-changes view for proposed text edits.
• Close the server when review is complete to avoid leaving a local port open.

Example use cases

• Author sends a draft email generated by the LLM for teammates to comment inline and suggest wording changes.
• Product writes a user-facing help article with the LLM, then uses the UI to collect editorial feedback and export a final changelog.
• Legal team reviews a policy summary produced by the model, annotates concerns, and exports structured feedback for revision.
• Instructor generates a model answer and collects student annotations and corrections directly in the document.
• PM collects stakeholder edits on a feature spec produced by the assistant and assembles a summarized change list.