wenhaochai

wenhaochai/claude-plugins

Plugin Claude Code

Plugin marketplace listing 3 plugins: anti-autoresearch, daily, writing.

wenhaochai/claude-plugins

Plugin Claude Code

Reviewer-side integrity forensics for autoresearch / AI-generated papers, used here as the ADVERSARIAL SELF-AUDIT gate on my own paper output (draft → /anti-autoresearch → fix verdict-bearing findings → re-run until CLEANGIVENEVIDENCE). 61 signals: 46 integrity hack-patterns in 8 families (numeric self-consistency, m.

wenhaochai/claude-plugins

Skill Claude CodeCodex

Synthesize the single strongest EVIDENCE-BOUND reviewer case to reject a paper, built ONLY from the evidence ledger (claims.json) + the other auditors' confirmed findings — never free-floating LLM critique. Two fresh cross-model codex threads: an attack writes the 200-word rejection paragraph (every accusation tagged…

wenhaochai/claude-plugins

Skill Claude CodeCodex

Transparent, itemized impressions of AI-generated WRITING STYLE — the repo's ONLY non-integrity track. Two passes: a deterministic defensive-hedge density screen (tools/checkaistyle.py, AIS-DEFENSIVE-HEDGE) plus a fresh cross-model GROSS-cases-only semantic pass over the 13 AIS- style tells (broken narrative arc, LLM…

wenhaochai/claude-plugins

Skill Claude CodeCodex

End-to-end substantive-integrity forensic sweep of a research paper (especially autoresearch / AI-Scientist-style output). Orchestrates the whole pipeline: ingest (arxiv-id | pdf | dir → working dir + pdftotext for L0) → /evidence-ledger (artifact manifest + observability level L0/L1/L2 + span-anchored claims.json) →…

wenhaochai/claude-plugins

Skill Claude CodeCodex

Audit whether a paper's baseline comparisons are COMPLETE, FAIR, and SIGNIFICANT: a required recent SOTA baseline is missing while 'best/SOTA' is claimed (HP-MISSING-BASELINE); a baseline is undertuned / given less compute-tuning-data, run at a mismatched config, or the equal-budget ablation-as-baseline is absent…

wenhaochai/claude-plugins

Skill Claude CodeCodex

Citation-integrity forensics: is every reference real, correctly attributed, and used in a context the cited work actually supports? Catches hallucinated references (no paper at the claimed arXiv id/DOI/venue, fabricated authors/year), metadata drift (wrong year/venue/version), and wrong-context citations (a real…

wenhaochai/claude-plugins

Skill Claude CodeCodex

Flagship intra-paper self-consistency forensics: does the paper contradict ITSELF across abstract/intro/tables/body/appendix, and does the method DESCRIBED match the method EVALUATED? Needs no external ground truth — works PDF-only (L0). Runs a deterministic arithmetic pass + a fresh cross-model semantic pass, every…

wenhaochai/claude-plugins

Skill Claude CodeCodex

Audit whether a paper's EVALUATION DESIGN actually measures what it claims and whether its reporting is complete — the validity layer family D (experiment-forensics) cannot reach. Three patterns: train/test leakage means the reported score may not measure generalization (HP-EVAL-LEAKAGE — adopts the Kapoor & Narayanan…

wenhaochai/claude-plugins

Skill Claude CodeCodex

Build the deterministic evidence ledger (artifactmanifest.json + claims.json) that every other Anti-Autoresearch auditor reads. One pass inventories artifacts, derives the observability level (L0 PDF-only / L1 +LaTeX / L2 +repo+results) by fixed rule, and extracts span-anchored, hashed, checkable claims (numbers…

wenhaochai/claude-plugins

Skill Claude CodeCodex

Audit experiment integrity against the evidence ledger. At L2 (repo + result files present) a fresh cross-model reviewer reads the eval code line-by-line for fake/derived ground truth, score self-normalization, phantom results (a paper number with no backing file/key), dead/uncalled metric code, verified-scope…

wenhaochai/claude-plugins

Skill Claude CodeCodex

MEMO-ONLY prior-work overlap advisory: surfaces the two ADVISORY taxonomy signals neither a tool nor a model can decide from the paper alone — ADV-TRIVIAL-COMBINATION (standard A+B+C / 缝合 stapling) and ADV-DUPLICATE-PUBLICATION (repackaged / duplicate submission). The executor RETRIEVES candidate prior work (DBLP…

wenhaochai/claude-plugins

Skill Claude CodeCodex

Checkable-ish surface presentation signals a reviewer notices first — duplicate/near-identical tables, leftover pipeline/template strings, too-few or LLM-looking figures, and page-padding. AUXILIARY ONLY and weak by design: a deterministic pass (tools/checkpresentation.py — dup-table + pipeline-artifact) plus a fresh…

wenhaochai/claude-plugins

Skill Claude CodeCodex

Family-G proof & derivation integrity forensics: does a THIRD PARTY's written proof/derivation actually establish its theorem, or does it skip an obligation, assume its own conclusion, take an invalid step, drift a symbol's meaning, or smuggle an unstated assumption? Decides from the WRITTEN proof/derivation …

wenhaochai/claude-plugins

Plugin Claude Code

Personal daily tools: /daily:todo (TODO.md with natural-language add/complete/delete/update) and /daily:spawn-remote (start another remote-control Claude Code in a detached tmux session over SSH).

wenhaochai/claude-plugins

Command

Spawn another remote-control Claude Code instance inside the current SSH session via detached tmux. Use when the parent CC is itself remote-controlled (cannot send TTY shortcuts) and the user wants a sibling CC URL to drive in parallel.

wenhaochai/claude-plugins

Command

Manage TODO.md (show / add / complete / delete / update via natural language) and show today's calendar.

wenhaochai/claude-plugins

Plugin Claude Code

Publication-prep tools for academic / public-facing writing. Bundles style (17 canonical English-prose rules + 18 page-cap additions + 2 audit-time rules), plot (matplotlib templates for paper figures), review (pre-submission self-review of own AI/ML paper Overleaf draft, simulating a top-tier ML venue reviewer…

wenhaochai/claude-plugins

Agent

Academic-paper editing for Overleaf-synced LaTeX projects (ML / CS / arxiv submissions). Invoke when the user wants to pull / edit / review / polish a paper section (abstract, intro, method, related work, experiments, conclusion, limitations); apply red-mark feedback wrapped in \textcolor{red}{...}; do bilingual zh/en…

wenhaochai/claude-plugins

Skill Claude CodeCodex

Matplotlib templates for paper / blog / report figures with a Google-brand palette, Palatino body font (matches arxiv mathpazo), and an announcement-clean default frame — L-shaped ink spines, no grid, left-aligned bold titles with a white-edged dot legend row above the axes (headerlegend + finalizeheaders), never…

wenhaochai/claude-plugins

Skill Claude CodeCodex

Pre-submission self-review of your own AI/ML paper Overleaf draft. Takes a project name (resolves to /Documents/docs/project/ /), recursively ingests the LaTeX source, and produces a reviewer-style critique markdown report simulating a top-tier ML venue reviewer (NeurIPS / ICML / ICLR / CVPR). Trigger on…

wenhaochai/claude-plugins

Skill Claude CodeCodex

Default writing standards. Use whenever producing English prose the user will send or publish: emails, message drafts, blog posts, tweets, documentation, grant proposals, and conference/journal paper drafts (NeurIPS, ICML, ICLR, ACL, CVPR, COLM, EMNLP, arxiv). Applies 17 canonical rules everywhere; 18 page-capped…

wenhaochai/codex-plugins

Skill Claude CodeCodex

Spawn a remote-control Claude Code instance from inside the current Codex session via detached tmux, and return its /remote-control URL. Use when the user wants to fire off a sibling Claude Code session they can drive in parallel from a browser (e.g. delegate a long task to CC while keeping the current Codex session…

wenhaochai/codex-plugins

Skill Claude CodeCodex

Manage TODO.md in the current project — show, add, complete, delete, update items via natural language. Use whenever the user mentions a TODO action ("加一条 X" / "X 完成了" / "删掉 X" / "X 改成 Y" / "add a todo" / "mark done" / "remove" / "check my todos"), references TODO.md, or asks to see today's calendar. Also handles…