interactive-reviewlisted
Install: claude install-skill anhermon/claude-public
## When to use this vs render-output
| Situation | Use |
|---|---|
| Output is final — just needs to be readable | `render-output` |
| Output needs user approval or changes | **this skill** |
| Eval results need grading | skill-creator eval viewer |
---
## Workflow
### Step 1 — Write the content
Write the plan, report, or analysis as markdown. Save to a temp file:
```
/tmp/<topic>-review.md
```
Use standard markdown. No special schema required (unlike render-output).
### Step 2 — Start the interactive viewer
```bash
python3 ~/.claude/skills/interactive-review/server.py \
/tmp/<topic>-review.md \
--workspace /tmp/<topic>-workspace \
--title "<Human-readable title>"
```
- Server starts on **port 3118** and opens the browser automatically.
- Feedback is saved to `/tmp/<topic>-workspace/feedback.json`.
- The page re-reads the markdown on each browser refresh — edit the file and refresh to see updates without restarting.
**With selectable options** (proposals, multi-choice decisions):
```bash
python3 ~/.claude/skills/interactive-review/server.py \
/tmp/<topic>-review.md \
--options /tmp/<topic>-options.json \
--preview-dir /tmp/<topic>-previews \ # optional: serve preview files at /preview/<file>
--workspace /tmp/<topic>-workspace \
--title "<Human-readable title>"
```
Options JSON format — each option can include a `preview_url` (relative `/preview/file.html` or absolute):
```json
[
{"id": "1", "label": "Short label", "description": "One-line expla