Webflow Mcp:Review Comments logo

Webflow Mcp:Review Comments

Organization
webflow
webflow-mcp:review-comments

Review open comment threads on a Webflow site and triage each one.

Overview

Publisherwebflow
Repositorywebflow-skills
Skill namewebflow-mcp:review-comments
Stars
122
Forks
18
Bundled files
Instructions only
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by webflow on GitHub. Read the source before you install it.

Installation

Install the Webflow Mcp:Review Comments AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/webflow/webflow-skills.git /tmp/webflow-skills
mkdir -p .claude/skills
cp -r /tmp/webflow-skills/plugins/webflow-skills/skills/review-comments .claude/skills/webflow-webflow-mcp-review-comments
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Webflow Mcp:Review Comments in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Webflow Mcp:Review Comments on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Webflow Mcp:Review Comments is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Review open comment threads on a Webflow site and triage each one.

Input: $ARGUMENTS — a site name (e.g. "Workhaus"), site ID (e.g. 6808fd4eff835ee3af009d6f), or either with the -reply flag (e.g. Workhaus -reply).

Flags:

  • -reply — in addition to writing the report, post a bot reply to each non-open thread. Without this flag, the skill runs in report-only mode (read-only).

Parse $ARGUMENTS at the start: strip -reply from the input to get the site identifier, and set replyMode = true if -reply was present, false otherwise.


Step 1 — Resolve the site

If the site identifier is empty or blank — cross-site comment survey

  1. Call data_sites_tool > list_sites to get all sites. Page through until all are collected.
  2. For each site, call data_comments_tool > list_comment_threads with isResolved: false and limit: 100. Page through until all unresolved threads are collected. If the API does not support isResolved filtering, fetch all threads and filter client-side to isResolved === false. Process sites in batches of 20 (batch multiple actions in a single tool call). After each batch completes, log a progress line: Batch {N}/{total} done (sites {start}–{end}): {summary of findings, e.g. "all 0 threads" or "SiteName has X threads, rest 0"}.
  3. For each site, compute:
    • unresolvedCount = total unresolved threads
    • newestDate = the maximum lastUpdated value across all unresolved threads for that site (ISO → human-readable date, e.g. "Apr 3, 2026"). If no threads, show .
    • oldestDate = the minimum lastUpdated value across all unresolved threads for that site (ISO → human-readable date). If no threads, show .
  4. Sort sites by unresolvedCount descending. Take the top 10.
  5. Above the table, show a heading line: ### Checked {totalSiteCount} sites — {sitesWithUnresolved} have unresolved comments.
  6. Display a table in this format (link just the site name to https://webflow.com/design/{siteId}):
| Site | Unresolved Comments | Newest / Oldest |
|------|---------------------|-----------------|
| [Site Name](https://webflow.com/design/{siteId}) ({siteId}) | {N} | {newestDate} / {oldestDate} |
  1. After the table, tell the user: "Run /webflow-mcp:review-comments <site name or ID> to review a specific site."
  2. Write the survey output to a file:
    • Ensure comment-reviews/ exists (create with mkdir comment-reviews if not).
    • Filename: comment-reviews/triage-report-{YYYY-MM-DD-HH-MM}.md using the current local time (zero-padded).
    • File contents: a bold H1 title # **Webflow Comment Review**, then a blank line, then an H3 line ### Report created on: {human-readable date, time, and timezone, e.g. "April 16, 2026 at 10:39 AM PDT"}, then a blank line, then the H3 heading line (### Checked …) and the full table from steps 5–6 above, in markdown.
    • Log the path after writing, e.g. Report written to comment-reviews/triage-report-2026-04-16-14-30.md.
  3. Stop — do not proceed to Step 2.

If the site identifier looks like a Webflow site ID (24-char hex), use it directly.

Otherwise call data_sites_tool > list_sites and find the site whose displayName matches the identifier (case-insensitive). If no match, tell the user and stop.


Step 2 — Fetch all open threads

Always make a fresh API call here — never reuse thread data from earlier in the conversation. The user may have added or resolved comments since the last run.

Call data_comments_tool > list_comment_threads with isResolved: false and limit: 100. Page through results until all threads are collected.

Log: Site: {displayName} and Found {N} open thread(s).


Step 3 — Build page-level element map

From the already-fetched thread list, build a frequency map of elementId.element → Set<pageId> across all threads. Any elementId.element that appears on 2 or more distinct pages is almost certainly the page root/body element (a real element ID would be page-scoped; only shared structural roots repeat across pages).

No API calls needed — this is a local computation on the thread data.


Step 4 — Triage each thread

For each thread:

4a — Fetch replies

Call data_comments_tool > list_comment_replies for this thread.

4b — Dedup check

Look for replies whose content includes the string — 🤖 Comment Review Agent.

If found, note the most recent one (lastAgentReply). If no human reply exists with a createdOn after lastAgentReply.createdOn, skip this thread (increment skipped count, continue to next thread).

4c — Compute element context

  • elementId = thread.elementId?.element
  • isPageLevel = elementId appears on 2 or more distinct pageIds in the frequency map from Step 3

4d — Classify the thread

Use the following criteria:

noise — No real design or engineering value:

  • Test/placeholder text ("hello world", "testing", "asdf", random characters)
  • Casual reactions with no ask ("looks nice", "nice!", "hey hey")
  • Duplicate sentiments that add nothing

stale — Real concern, but old and likely handled:

  • Substantive comments older than 14 days with no replies and no follow-up
  • Questions that are probably resolved ("beta for how long?", "is this good contrast?")
  • Action items that normal review would have caught

open — Real, actionable concern needing attention:

  • Explicit tasks ("Should be sentence case", "fix image", "Look at name wrt L10N")
  • Design decisions still required
  • Specific and concrete concerns

page-level — Comment is on the page root, not a specific element:

  • Use this when isPageLevel is true
  • The comment is not anchored to any specific element — it may be intentional or may be an orphan from a deleted element

4e — Compose reply

Always compose the reply text (it appears in the report regardless of mode):

  • noise: one sentence confirming it's safe to resolve. E.g. "Looks like test text — safe to resolve."
  • stale: state the age in days, suggest resolving, invite reopen. E.g. "This is 302 days old with no follow-up. Safe to resolve — reply here if it's still relevant."
  • open: no reply text — surface in report only.
  • page-level: one sentence noting it's not attached to a specific element. E.g. "This comment is on the page root rather than a specific element — it may be an orphan from a deleted element. Safe to resolve if no longer relevant."

Append \n\n— 🤖 Comment Review Agent to every reply text.

4f — Post reply (only if replyMode = true)

If replyMode is true and verdict is not open, call data_comments_tool > create_reply with the composed reply content.

Log each thread as:

{VERDICT_EMOJI} {verdict}  "{preview (60 chars)}"
                {one-sentence classification reason}
                ↳ {reply posted | no reply — report only | skipped}

Verdict emojis: noise = 🗑, stale = 🕰, open = 🔴, page-level = 📄


Step 5 — Write the report

After processing all threads:

  1. Check whether a comment-reviews/ directory exists at the top level of the working directory. If it does not exist, create it with mkdir comment-reviews.
  2. Write a markdown report to comment-reviews/{slugified-site-name}-comments-triage-report.md (lowercase, hyphens, no special chars).

Report format:

markdown
# Comment Review — [{siteName} ({siteId})](https://webflow.com/design/{siteId})

**Run:** {human-readable date and time}
**Mode:** {Report only | Report + replies posted}
**Threads:** {total} total | 🔴 {open} open | 🕰 {stale} stale | 📄 {page-level} page-level | 🗑 {noise} noise | ⏭ {skipped} skipped

---

## 🔴 Needs Attention ({count})

| Comment | Author | Age | Link |
|---------|--------|-----|------|
| "{first 80 chars of content}" | {author.name} | {age in days}d | [Open ↗]({thread.url}) |

## 🕰 Stale — Candidates to Resolve ({count})

| Comment | Author | Age | Suggested Reply | Link |
|---------|--------|-----|-----------------|------|
| "{first 80 chars of content}" | {author.name} | {age in days}d | {composed reply text, without the `— 🤖 Comment Review Agent` suffix} | [Open ↗]({thread.url}) |

## 📄 Page-level — Not Anchored to a Specific Element ({count})

| Comment | Author | Age | Suggested Reply | Link |
|---------|--------|-----|-----------------|------|
...

---

**🗑 Noise:** {count} thread(s) — {if replyMode: "replied to" | if report-only: "suggested replies in report"}. Safe to bulk-resolve.

Use _None._ for any section with no entries.

Log a summary line and confirm the report path.

Frequently asked questions

What does the Webflow Mcp:Review Comments AI skill do?

Review open comment threads on a Webflow site and triage each one.

Why use Webflow Mcp:Review Comments on TypingMind?

Because you install it once and use it with any model. Webflow Mcp:Review Comments is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Webflow Mcp:Review Comments in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/webflow/webflow-skills/tree/main/plugins/webflow-skills/skills/review-comments. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Webflow Mcp:Review Comments?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Webflow Mcp:Review Comments?

As many as you like. As long as a model supports skills, you can use Webflow Mcp:Review Comments with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Webflow Mcp:Review Comments AI skill free?

Yes. It is published on GitHub by webflow under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇