菜单
Bulk-post Overleaf comments from a JSON plan. Use when asked to add many Overleaf comments at once (e.g. thesis review, paper feedback, multi-line annotations). Drives an existing logged-in browser tab via the chrome-cdp skill.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
Multi-agent review-and-improve loop for a GitHub PR you have checked out — posts a "starting" PR comment cc'ing the original author, runs requested rounds plus any adaptive continuation, applies fix commits to the local branch after each round, pushes everything back to the PR, then edits the starting comment in-place with the synthesized report (or a failure summary). Auto-detects the PR from the currently checked-out branch when no locator is supplied. Use when the user wants to "improve a PR", "review and commit fixes", "iterate on my PR", or "review and push back" against a checked-out PR branch. Requires `gh`, `uuidgen`, `jq`, and `uv` or `python3` on PATH. Activates the `review-anvil` engine in per_fix mode.
Read-only multi-agent review of a GitHub Pull Request, with the synthesized report posted back as a PR comment so the author is notified. Use when the user wants to review a GitHub PR (github.com or GitHub Enterprise) and post a structured review back to the PR conversation. Auto-detects the PR from the currently checked-out branch when no locator is supplied. Requires `gh`, `uuidgen`, `jq`, and `uv` or `python3` on PATH. Activates the `review-anvil` engine in read-only mode and orchestrates the shell helper for posting.
Multi-agent code review without any edits or commits — requested rounds of parallel reviewers, synthesis, default-on reproduction of uncertain material findings, and a final report. Use when the user wants a code review of a diff, branch, path, or PR but does NOT want any fixes applied automatically (no edits, no commits, no staging). Activates the `review-anvil` engine in read-only mode.
Iteratively refine code via requested rounds of parallel subagent review and orchestrator-applied fixes, with bounded adaptive continuation enabled by default for productive runs. Use when the user says "let's do three rounds", "fix/review loop", "back-and-forth review", "iterative review", or asks to harden a change with multiple rounds of codex/claude review.
Convert academic PDFs to clean Markdown. Use whenever you encounter a PDF paper (local file or URL) that needs to be read, analyzed, or referenced as text.
Delegate code review, plan review, and exploration to Codex CLI. Use when you need an independent second opinion on code changes, plans, or architecture — or when asked to run codex review/exec.
| name | overleaf-comment |
| description | Bulk-post Overleaf comments from a JSON plan. Use when asked to add many Overleaf comments at once (e.g. thesis review, paper feedback, multi-line annotations). Drives an existing logged-in browser tab via the chrome-cdp skill. |
Post N Overleaf comments from a JSON plan, one CDP eval per item. Uses the user's already-logged-in browser session via the chrome-cdp skill.
Break-even is around 5 comments — for fewer, type them directly in Overleaf.
chrome-cdp skill installed; the wrapper auto-discovers cdp.mjs (or set $CDP_BIN).--remote-debugging-port=9222.# 1. find the project tab
${CLAUDE_PLUGIN_ROOT}/skills/overleaf-comment/overleaf-comment --list
# 2. post a plan (target prefix from `cdp list`)
${CLAUDE_PLUGIN_ROOT}/skills/overleaf-comment/overleaf-comment 6172EEAA plan.json
# or supply the project URL — the wrapper picks the matching tab
${CLAUDE_PLUGIN_ROOT}/skills/overleaf-comment/overleaf-comment --url https://www.overleaf.com/project/<id> plan.json
For self-hosted Overleaf set OVERLEAF_HOSTS="overleaf.example.com" (space-separated for multiple).
A JSON array of {file, line, text}:
[
{"file": "main.tex", "line": 8, "text": "linespread should be 1.5"},
{"file": "parts/200-discussion.tex", "line": 108, "text": "ungrammatical"}
]
file matches Overleaf's file tree by basename (or path-suffix when the plan supplies /).line is 1-indexed. Blank lines cannot be commented on.text is the comment body; multi-line is fine via \n in the JSON string.Per-item results stream to stdout:
[main.tex] OK: main.tex (155 lines)
L8: OK: line 8
L109: FAIL: line 109 is blank — Overleaf cannot anchor a comment
A summary, identity preflight, and any FAIL diagnostics go to stderr.
Exit codes: 0 = all OK · 1 = at least one FAIL/SKIP · 2 = usage / validation · 75 = EX_TEMPFAIL (background tab, "Out of sync" modal — re-run after fixing).
Residual plan: if any items failed or were skipped, the wrapper writes <plan>.residual.json containing just those items. Retry with overleaf-comment <target> <plan>.residual.json.
page-helpers.js for easy re-targeting on a redesign.posting as Overleaf account: <email> before running so you can verify.Comments land on any file Overleaf renders in CodeMirror — including .tex, .bib, .bst, .cls, .sty. The skill auto-selects one of two paths:
.tex): selects the line and clicks the static "Add comment" button..bib): Overleaf suppresses the toolbar button for non-tex modes but still mounts a floating Add-comment popover on a trusted PointerEvent-driven selection. The wrapper synthesises a real mouse drag via CDP Input.dispatchMouseEvent to trigger the popover, then clicks through. The fallback fires automatically when the toolbar button is absent — no plan changes required.