version: "1.0.1" name: review-blog description: Use when user says "review blog", "review this blog", or /review-blog. Uploads Markdown or Google Docs URL for annotation, waits for comments, synthesizes structured feedback, publishes as a shareable Google Doc. argument-hint: <file.md or Google Docs URL>

Review Blog

Overview

End-to-end blog review: upload to Google Docs, wait for annotations, synthesize feedback, publish as a shareable doc.

Core Workflow

Step 1 — Prepare

Goal: Get the article into Google Docs for annotation.

Exit criterion: Google Docs URL presented to the user.

Extract the file path or URL from the skill argument. If no argument, ask the user.

Run review-blog-start <input> and parse the JSON output. The url field contains the Google Docs URL.

Present the URL to the user:

Here's the doc: {url}

Go annotate it with your comments. Tell me when you're done.

Stop and wait for the user.

Step 2 — Wait

Goal: User confirms they're done annotating.

Exit criterion: User said they're done (or equivalent).

Do nothing until the user comes back. Any indication that comments are ready counts: "done", "finished", "go", "comments are in", etc.

Step 3 — Fetch

Goal: Retrieve the article content and comments from Google Docs.

Exit criterion: Markdown content and comments JSON in hand.

Run both commands, capturing their output:

1. gdoc2md <url> — outputs a directory path. Read {outputDir}/index.md for the article markdown.
2. gdoc-comments-json <url> — outputs a JSON array of { anchor, comment } objects.

Step 4 — Synthesize

Goal: Cross-reference comments against the full article and produce structured feedback.

Exit criterion: Feedback markdown written to a temp file.

1. Read the article markdown and the comments JSON.
2. Read the output format reference at references/output-format.md.
3. Auto-detect the article's language (default to English).
4. For each comment:

• Match its anchor to the corresponding passage in the article.
• If the issue raised by the comment is addressed later in the article, discard it (auto-resolve).
• If the comment is in French but the article is in another language, translate the feedback.

1. Group remaining comments by thematic axis (Narration, Content, Clarity/Style, Diagrams, etc. — axes adapt to the article). Within each axis, order from most impactful to least.
2. Write the feedback following the output format reference: TL;DR, Strengths, Improvements grouped by axis.
3. Write the feedback markdown to /tmp/oroshi/review-blog/feedback.md.

Step 5 — Publish

Goal: Upload the feedback as a shareable Google Doc.

Exit criterion: Shareable URL returned to the user.

Run md2gdoc --title "<Article Title> (review)" --no-open /tmp/oroshi/review-blog/feedback.md. The command outputs the Google Docs URL.

Present the URL to the user:

Here's the review: {url}

Common Rationalizations

Checklist

• [ ] Input resolved (file path or URL)
• [ ] review-blog-start called and URL presented
• [ ] Waited for user confirmation
• [ ] gdoc2md called and article markdown read
• [ ] gdoc-comments-json called and JSON parsed
• [ ] Output format reference read
• [ ] Article language detected
• [ ] Comments cross-referenced against full article
• [ ] Addressed comments auto-resolved
• [ ] French comments translated to article language
• [ ] Feedback written following output format (TL;DR, Strengths, Improvements)
• [ ] Feedback published via md2gdoc --title and URL returned