plannotator_skill

This skill helps you review and annotate AI agent plans and diffs with a visual UI, enabling structured feedback and auto-save.
  • Shell

24

GitHub Stars

2

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill supercent-io/skills-template --skill plannotator

  • SKILL.md17.0 KB
  • SKILL.toon2.9 KB

Overview

This skill provides an interactive visual UI to review and annotate AI agent plans and git diffs before or after code execution. Use it to approve, request changes, or send structured feedback back to the agent, and optionally auto-save approved plans to Obsidian or Bear Notes. It integrates with Claude Code and OpenCode and supports image annotation and team sharing links.

How this skill works

When an agent finishes planning or produces a diff, the UI displays the plan or unified/split diff for visual annotation. You mark steps or line ranges with structured intents (delete, insert, replace, comment), then submit a single outcome: approve or request changes. Annotations are sent back to the agent as structured feedback; approved plans can be auto-saved to note apps. The CLI includes scripts to install, configure hooks or plugins, run reviews, and set remote/devcontainer modes.

When to use it

  • Review an agents implementation plan before it starts coding
  • Annotate and give feedback on a generated git diff after coding
  • Create a feedback loop to request concrete replanning from the agent
  • Share a plan review with teammates via a link for collaborative sign-off
  • Auto-save approved plans to Obsidian or Bear Notes for documentation
  • Run reviews from remote environments (SSH, devcontainer, WSL) without opening a browser locally

Best practices

  • Run plan review before implementation to catch design or safety issues early
  • Keep each annotation targeted to one concrete, actionable change
  • Include acceptance criteria when requesting changes so the agent can replan accurately
  • Annotate exact line ranges in diffs and attach screenshots for UI/UX feedback
  • Use the CLI status and hook checks to verify integration before relying on automatic triggers

Example use cases

  • Trigger automatic plan review when exiting Claude Code plan mode to block risky implementations
  • Launch a visual diff review for all uncommitted changes or a specific commit
  • Configure remote mode for a headless server and forward the UI to your local browser
  • Save approved implementation plans into an Obsidian vault for team documentation
  • Use image annotation to mark UI issues in a feature branch before merge

FAQ

Install the CLI and configure the editor hook or install the Claude Code/OpenCode plugin; the hook opens the review UI when the agent exits plan mode.

Can I review a specific commit or branch diff?

Yes. The review script accepts refs (for example HEAD~1 or main...HEAD) and opens a diff review UI where you can annotate lines and submit feedback.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational