[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-80784":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":14,"contributorsCount":14,"subscribersCount":14,"size":14,"stars1d":15,"stars7d":15,"stars30d":15,"stars90d":14,"forks30d":14,"starsTrendScore":16,"compositeScore":17,"rankGlobal":9,"rankLanguage":9,"license":18,"archived":19,"fork":19,"defaultBranch":20,"hasWiki":21,"hasPages":19,"topics":22,"createdAt":9,"pushedAt":9,"updatedAt":23,"readmeContent":24,"aiSummary":25,"trendingCount":14,"starSnapshotCount":14,"syncStatus":15,"lastSyncTime":26,"discoverSource":27},80784,"ai-peer-review-skill","AlexWortega\u002Fai-peer-review-skill","AlexWortega","Claude Code skill for multi-reviewer peer review of academic papers. Adapted from poldrack\u002Fai-peer-review — uses parallel Claude subagents instead of multiple proprietary LLMs.",null,"Python",41,3,39,0,2,6,47.01,"MIT License",false,"main",true,[],"2026-06-12 04:01:30","# ai-peer-review-skill\n\nA [Claude Code](https:\u002F\u002Fclaude.com\u002Fclaude-code) skill that runs a multi-reviewer peer review of an academic paper.\n\nDrop a PDF in, get back N independent reviews + a synthesized meta-review + a CSV table of which reviewer raised which concern.\n\n## What it does\n\nGiven a PDF (or DOCX \u002F `.txt` \u002F `.md`) of a paper, the skill:\n\n1. Extracts the text.\n2. **Spawns N reviewer subagents in parallel** with anonymized NATO codenames (`alfa`, `bravo`, `charlie`, …). Each subagent sees only the paper and produces an independent, structured review (summary → major concerns → minor concerns → verdict).\n   - By default, one of the panel slots is filled by an **AI Alignment Forum-style critic** that follows Neel Nanda's *[Highly Opinionated Advice on How to Write ML Papers](https:\u002F\u002Fwww.alignmentforum.org\u002Fposts\u002FeJGptPbbFPZGLpjsp\u002Fhighly-opinionated-advice-on-how-to-write-ml-papers)* — hard red-teaming on narrative, novelty, baselines, ablations, post-hoc analysis, p-value rigor, reproducibility, and an explicit \"what did this update in my beliefs?\" check. Disable with `alignment_critic=false`.\n3. **Synthesizes a meta-review** in the main thread, identifying common vs unique concerns, ranking the reviewers by usefulness, and producing a final verdict.\n4. **Extracts a concerns table** — a boolean matrix of `concern × reviewer` — and saves it as CSV.\n5. Bundles everything into `results.json`.\n\nOutput layout:\n\n```\npapers\u002F\u003Cpaper-stem>\u002F\n├── review_alfa.md\n├── review_bravo.md\n├── review_charlie.md\n├── review_delta.md\n├── review_echo.md\n├── meta_review.md\n├── concerns_table.csv\n└── results.json\n```\n\n## Where it came from\n\nThis is a Claude Code skill port of [**poldrack\u002Fai-peer-review**](https:\u002F\u002Fgithub.com\u002Fpoldrack\u002Fai-peer-review) by Russ Poldrack — a Python tool that calls 6 different proprietary LLMs (GPT-4o, GPT-4o-mini, Claude 3.7 Sonnet, Gemini 2.5 Pro, DeepSeek R1, Llama 4 Maverick) to peer-review a paper, then synthesizes a meta-review.\n\nThe port differs in two ways:\n\n| | Original | This skill |\n|---|---|---|\n| Reviewers | 6 different proprietary LLMs | N parallel Claude subagents (default 5) |\n| API keys needed | OpenAI + Anthropic + Google + Together | None beyond Claude Code itself |\n| Diversity | True cross-model diversity | Independent generations, single model family |\n| Domain | Hard-coded to neuroscience | `domain` argument, inferred from paper if omitted |\n\nThe skill keeps the original's artifact layout (`review_*.md`, `meta_review.md`, `concerns_table.csv`, `results.json`) and the NATO-codename anonymization scheme so outputs are interchangeable.\n\nIf you actually need cross-model diversity (e.g. for a methods paper *about* AI peer review), use the original `poldrack\u002Fai-peer-review` Python tool instead. The SKILL.md documents this fallback explicitly.\n\n## Install\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FAlexWortega\u002Fai-peer-review-skill.git\nln -s \"$(pwd)\u002Fai-peer-review-skill\" ~\u002F.claude\u002Fskills\u002Fpaper-review\n```\n\n(or symlink into `\u003Cproject>\u002F.claude\u002Fskills\u002Fpaper-review` for project-scoped install.)\n\nRestart your Claude Code session so the skill is picked up.\n\n## Use\n\nIn Claude Code:\n\n> Peer-review this paper: `~\u002FDownloads\u002Fmanuscript.pdf`\n\nor\n\n> \u002Fpaper-review ~\u002FDownloads\u002Fmanuscript.pdf\n\nOptional knobs (just say them in plain language):\n\n- domain — `\"neuroscience and brain imaging\"`, `\"reinforcement learning\"`, etc.\n- num_reviewers — 3 to 8 (default 5)\n- output_dir — defaults to `.\u002Fpapers\u002F\u003Cpaper-stem>\u002F`\n- skip_meta — only individual reviews, no synthesis\n- overwrite — regenerate existing `review_*.md` files\n\n## Layout\n\n```\n.\n├── SKILL.md                              # frontmatter + workflow Claude follows\n├── prompts\u002F\n│   ├── reviewer.md                       # generic reviewer template\n│   ├── reviewer_alignment_forum.md       # AAF-style critic (Nanda framework)\n│   └── metareview.md                     # synthesis template\n└── README.md                             # this file\n```\n\n## Credits\n\n- Original tool and prompt design: [Russell Poldrack](https:\u002F\u002Fgithub.com\u002Fpoldrack) — [poldrack\u002Fai-peer-review](https:\u002F\u002Fgithub.com\u002Fpoldrack\u002Fai-peer-review)\n- Skill adaptation: this repo\n\n## License\n\nMIT (skill adaptation only). The upstream `poldrack\u002Fai-peer-review` repo is unlicensed at the time of this port; only the design and workflow are referenced here, no upstream code is redistributed.\n","ai-peer-review-skill 是一个用于学术论文多审稿人同行评审的 Claude Code 技能。该项目通过并行生成多个 Claude 子代理来独立审查论文，每个子代理提供结构化的评审报告，包括摘要、主要和次要问题以及结论。此外，项目还包含一个 AI 对齐论坛风格的批评者，可以对论文进行严格的红队测试，并在最后合成一份元评审报告及问题表。适合需要高效、自动化的学术论文评审场景，尤其是当用户希望减少对多种专有语言模型依赖时。使用 Python 编写，遵循 MIT 许可协议。","2026-06-11 04:02:18","CREATED_QUERY"]