# Review AI Plans and Diffs Visually

AI coding agents can start work before plans are clear, which makes errors costly. Plannotator adds a browser review step for plans, diffs, images, approvals, and structured feedback.

## Install

```bash
npx skillstore add supercent-io/plannotator
```

## Metadata

- Status: approved
- Slug: supercent-io-plannotator
- Version: 0.9.2
- Author version: 0.9.2
- Skillstore revision: r1
- Version status: valid
- Tree hash: 260b1369c309ad31d64887e7bbb5b317d817735493a1c2789af8de60a98dd5ef
- Author: supercent-io
- GitHub username: supercent-io
- License: MIT
- Repository: https://github.com/supercent-io/skills-template/tree/main/.agent-skills/plannotator/
- Ref: 3e4b6c31a74a3bd1a291c98cf585d720cb9fbc88
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: blocked
- Manual install advisory: allowed\_with\_warning
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands, network, filesystem
- Quality score: 38
- Quality tier: warning
- Public page: https://skillstore.pages.dev/skills/supercent-io-plannotator
- Manifest: https://skillstore.pages.dev/api/skills/supercent-io-plannotator/manifest

## Capabilities

- Installs the plannotator CLI and can configure integrations for multiple AI coding tools.
- Adds plan review hooks for Claude Code and Gemini CLI when the setup scripts are run.
- Creates Codex prompt and developer instruction files for manual plan review workflows.
- Registers OpenCode plugin settings and slash commands for review and annotation.
- Launches a visual git diff review workflow through the plannotator review command.
- Checks local integration status for hooks, prompts, environment variables, and git changes.

## Use Cases

- Review an Agent Plan Before Coding: Open a proposed implementation plan in a browser, annotate risky steps, and send structured feedback before code changes start.
- Inspect a Git Diff With Line Notes: Launch a visual diff review for uncommitted work or a selected commit range, then capture focused feedback for revisions.
- Coordinate Reviews Across Agent Tools: Configure similar plan review flows for Claude Code, Codex, Gemini CLI, and OpenCode in one local workflow.

## Prompt Templates

### Start a Basic Plan Review

```
Use plannotator to review my implementation plan before coding. Open the plan review flow and wait for my approval or requested changes.
```

### Review Current Git Changes

```
Open a plannotator diff review for my current git changes. Summarize the review feedback and update the code only after I approve the changes.
```

### Configure an Agent Integration

```
Help me configure plannotator for my AI coding tool. Show the files that will change first, then apply the setup only after I confirm.
```

### Run a Full Review Workflow

```
Create an implementation plan, send it to plannotator for review, revise it from my annotations, then implement only after approval. After implementation, open a diff review and address approved feedback.
```

## Limitations

- The installer depends on external plannotator.ai scripts and should be verified before use.
- Setup scripts modify hidden AI tool configuration files in the user home directory.
- Browser review, notes export, and diff review require the plannotator CLI to be installed locally.
- The skill supports review workflows, but it does not replace human technical judgment.

## Best Practices

- Run dry-run modes before setup so hidden configuration changes are visible.
- Review plans before implementation and include acceptance criteria in requested changes.
- Verify installed scripts and backups before enabling hooks across multiple AI tools.

## Anti Patterns

- Do not pipe remote installers into a shell without independent verification.
- Do not enable every integration before confirming which tools your team uses.
- Do not place sensitive plans in shared temporary paths on multi-user systems.

## Security Audit

- Audited at: 2026-07-09T09:18:05.64\+00:00
- Summary: The skill provides a legitimate visual review workflow, but it has high security impact because installation and setup scripts execute or encourage remote code execution and persistently modify agent configuration. Many Markdown backtick and read-only status findings are false positives, while the installer, hidden configuration writes, temporary plan files, and shell profile editing remain confirmed risks.

## Stats

- Views: 169
- Downloads: 12
- Favorites: 0
- Popularity score: 0