aidlc-io/aidlc/annotron-review
Claude Code and agent skill by aidlc-io/aidlc
- community
Checking that the SKILL.md file is still in the repository.
Open agent-generated HTML artifacts in a local browser editor so the user can
Install
npx vskill@latest install aidlc-io/aidlc/annotron-reviewSKILL.md
First-time setup
Before running annotron for the first time in a project, check if .claude/settings.json already allows Bash(curl:*) and Bash(annotron:*). If not, ask the user:
"annotron cần tự động chạycurlvàannotron pollđể nhận feedback mà không cần bạnapprove từng lần. Cho phép thêm auto-allow vào.claude/settings.jsonkhông?"
If yes, create/update .claude/settings.json:
When the user wants to review or refine an HTML artifact you produced:
- Open it: run
annotron <path-to-artifact.html>.
This starts the background server (if not running), registers the file, and opens the review editor in the user's browser. Print the editor URL for the user.
- Wait for feedback: run
annotron poll <path>.
This blocks until the user sends feedback. Output is JSON with:
items[]— each haskind(element | text),selector,text,note, and an
optional images[] (images attached to that specific annotation's note/reply box)
message— freeform message from the userimages[]— top-level images attached to the composer message (or a single annotation reply)- Every image (top-level or per-item) has
nameandpath(an absolute file path).
Read each path to see the image before applying the feedback.
finalized: true— signals the user is done; skip to step 5
- Apply the feedback: edit the HTML file using the selector/text + note from each item.
Each item in items[] now includes a server-assigned id field (e.g. ann_xxx) you can use to reply to a specific annotation thread.
Your steps are mirrored automatically. A bundled PostToolUse/PreToolUse hook streams every tool call (Read/Edit/Bash/…) into the browser sidebar as CLI-like steps, and reflects idle/permission status — you do not need to narrate. You may still post a high-level milestone with annotron progress <path> "…" if it helps, but it is optional now.
Remote permission approval (optional, user-controlled). If the user turns on Remote approve in the browser, the hook routes each tool's permission prompt to the browser; the user clicks Allow / Allow-always / Deny and the decision comes back to the CLI. If a tool is denied you'll get a deny reason — respect it, don't retry the same action. If the browser doesn't answer in time the prompt falls back to the terminal. This is transparent to you; no special commands needed.
Cancellation is enforced automatically. annotron ships a PreToolUse hook: if the user clicks Cancel while you are mid-round, your next tool call (Edit/Write/Read/Bash, except annotron commands) is denied with a message telling you to stop. When you see that denial, do not retry other edits — reply with what you did and did not finish (annotron poll <path> --reply "Stopped — cancelled. Applied X, skipped Y.") and wait for the next feedback. The cancel flag resets automatically when the user sends new feedback. (You can also proactively annotron check <path> between steps; it prints {"cancelled":true} — but the hook makes this optional.)
- Tell the user what changed: run
annotron poll <path> --reply "...".
This posts an agent message to the general conversation log, then re-arms the poll. To reply directly to a specific annotation thread, pass --annotation-id: annotron poll <path> --reply "message" --annotation-id ann_xxx The browser shows that reply inline in the annotation's thread. A reply without --annotation-id posts to the general conversation log. Saving the file triggers an automatic live-reload in the browser.
- Repeat steps 2–4 until the poll returns
"finalized": true.
At that point the user has written the confirmed result into the file — the loop is done.
Tips
- The
bin/directory containingannotronis on your PATH while this plugin is active. - Keep replies short and action-focused ("Updated h1 colour, fixed revenue figure, added footer.").
- Never edit the file while a poll is in flight — wait for the feedback JSON first.
- Annotations are persisted automatically in
<artifact>.annotron.jsonbeside the HTML file. - Pasted/uploaded images are saved to
.annotron-uploads/beside the artifact; feedback references them by absolutepath—Readthose paths to view the images. - Post progress liberally with
annotron progress <file> "…"— one short line per step. Keep it action-focused ("Reading X", "Updating Y"). annotron check <file>is cheap; run it between edits so a mid-flight cancel is honored quickly.- To reply to a specific annotation thread:
annotron poll <file> --reply "msg" --annotation-id <id>— the browser shows the reply inline in that annotation's thread.
Security
- Security scanPassed, 93 of 100Pattern checks on the skill's files.
- Model reviewNot runA model review is scheduled for the next rescan.
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingvendor/annotron/skills/annotron-review/SKILL.md
Popularity
- Repository stars
- 56
- Repository forks
- 30
- 7-day trend score
- 20
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/aidlc-io/aidlc/annotron-review)