Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add Nero1688/claude-academic-skills --skill thesis-consistency-auditgit clone --depth 1 https://github.com/Nero1688/claude-academic-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/nero1688/claude-academic-skills/thesis-consistency-audit)<a href="https://agentmods.dev/skills/nero1688/claude-academic-skills/thesis-consistency-audit"><img src="https://agentmods.dev/badge/skills/nero1688/claude-academic-skills/thesis-consistency-audit/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/nero1688/claude-academic-skills/thesis-consistency-audit"><img src="https://agentmods.dev/badge/skills/nero1688/claude-academic-skills/thesis-consistency-audit.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00500 | $0.03040 |
| Opus 5 | $0.00250 | $0.01520 |
| Sonnet 5 | $0.00100 | $0.00608 |
| Haiku 4.5 | $0.00050 | $0.00304 |
Grade A, and why
thesis-consistency-audit scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 12d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
階段一:機械對帳(若有 .docx)
scripts/audit_docx.py 讀取 .docx,自動做機械層對帳:表格加總、跨表樣本數一致性、不可能值與離群、敘述↔迴歸 N 落差。
執行:python scripts/audit_docx.py 論文.docx
(此腳本僅依賴 python-docx。若使用者只貼文字/表格而非 .docx,跳過腳本,全部改人工,並在報告註明「未跑機械對帳」。)
階段一之二:雙盲投稿的身分資訊清除(要投雙盲期刊才做)
scripts/anonymize_office.py 稽核並清除 .docx/.pptx/.xlsx 內藏的作者身分:
python scripts/anonymize_office.py 論文.docx # 只稽核,不動檔
python scripts/anonymize_office.py 論文.docx --apply # 匿名化(雙盲)
python scripts/anonymize_office.py 論文.docx --apply --strip-comments --strip-revisions
會抓出五個藏身處:core.xml 建立者/最後修改者、app.xml 機構與主管、
custom.xml 自訂屬性(常有計畫編號)、comments.xml 每則註解的作者名、
以及追蹤修訂的 w:author。Word 內建的「檢查文件」不一定清得掉後兩者。
⚠️ 紀律:註解與追蹤修訂含有內容,不只是中繼資料,故預設只報告不刪除,
要刪必須明確加旗標;清除前一律自動備份。
⚠️ 雙盲投稿請用預設匿名模式,不要用 --set-author(那是給最終定稿/存檔版)。
⚠️ 本工具只處理檔案中繼資料。正文自我引用(「作者先前研究(陳,2024)」)、
致謝、基金計畫編號、檔名含姓名——這些才是最常見的雙盲破功點,必須人工檢查。
階段二:六維度人工核對
機械層之外,依 references/audit_checklist.md 逐項核對。何時讀該檔:要展開任一維度的細目 checkbox 時讀它;本頁只列維度綱要。每項列出「衝突在哪(具體數字)+怎麼改」。
-
假設↔迴歸表對齊:每個假設(H1/H2/H3)由哪張表、哪個應變數檢定?表註標的 H 編號正確嗎?係數方向與顯著性是否與假設陳述一致?有沒有把 A 應變數的假設標在 B 應變數的表上?
-
樣本數一致性:樣本篩選表最終 N → 敘述統計 N → 迴歸觀察值,三處是否一致?落差是否以一句話交代(落後期、listwise 刪除)?敘述統計是否與迴歸用同一樣本?
-
篩選聲明↔敘述統計:篩選表若寫「剔除遺漏值(0)」,敘述統計各變數 N 是否真的都等於全樣本?有變數 N 較少卻宣稱 0 遺漏即為矛盾。
-
資料品質界限:不可能值(董事會規模最小=0、比率超出 [0,1] 或 [0,100])?極端離群(max 偏離 mean 逾~10 SD)?連續變數是否縮尾並註明?
- 相關 >0.9 紅旗(L-002):相關矩陣中任兩變數 |r|>0.9 → 疑似共線或「同一變數誤當兩個放入」(平方項未置中、虛擬變數代錯,如 RPT 誤代 TNFD)。要作者查是否誤代或應置中,不可放著。
- 假說→變數→資料欄位三欄對照(L-002):每個假設用到的變數,對到的 TEJ 欄位/代碼是否正確?
-
文字↔表格:內文引述的樣本期間、比例、係數、N、轉折點等,是否與表格逐一吻合?交乘項/二次項的邊際效果與轉折點判讀是否與係數一致(轉折點 = -β₁/β₃ 或 -β₁/(2β₂))?
- 同值重複紅旗(L-003):全表掃「不同欄格出現完全相同數值」(如兩個不同變數的 t 值都是 38.617)→ 疑似複製貼上貼錯,回溯每個數字的來源輸出檔位置。
-
引用與表註一致:內文 APA 引用 ↔ 參考文獻清單(孤兒引用、漏列)?顯著性星號(//)是否在表註定義?表註的應變數說明與該表實際一致? (引用只做內文↔清單雙向對帳;文獻是否真實存在/幻覺,轉 citation-verifier 無網稽核或 check-citations 有網查驗——Claude Code 環境須加 anthropic-skills: 前綴。)
<output_contract> 結論先行。報告依嚴重度分區,每項附證據錨點(表號、頁、欄格)。嚴重度三級:
- Fatal=會直接被審查委員抓、且動搖結果可信度(如變數代錯導致主結果、樣本數對不上且無法解釋)。
- Major=需重跑分析或補說明才能修(離群未縮尾、N 落差未交代)。
- Minor=可直接改的文字/註記不一致(表註漏定義星號、內文年份筆誤)。
# 一致性稽核報告:<論文>
## 機械對帳結果
(貼 audit_docx.py 輸出摘要;或註明「未跑機械對帳,全人工」)
## 待修清單(依嚴重度)
[Fatal] <維度N> <衝突在哪,具體數字> | 錨點:表X/頁Y | 怎麼改:<...>
[Major] ...
[Minor] ...
## 需重跑分析 vs 可直接改文字(明確分開)
- 需原始資料重跑:<...>(標「待作者以原始資料修正」)
- 可直接改文字/註記:<...>
## 寫對的地方(不必動)
<...>
</output_contract>
輸出節錄:
- [Fatal] 維度5 同值重複(Table 3, Model 1):家族持股 t=7.373 與董事會獨立性 t=7.373 完全相同,統計上兩個不同變數 t 值相同機率極低,疑複製貼上貼錯(對照 L-003:真值可能一為 7.373、一為 38.617 類誤植)。怎麼改:回原始 regression output 逐欄核對,指出正確值來源列。標「待作者以輸出檔修正」。
- [Fatal] 維度4 相關>0.9(相關矩陣):家族持股與家族控制 r=0.98,疑為同一構念以兩種代理重複放入模型,或變數代錯(L-002 型)。怎麼改:確認兩者操作型定義是否實質不同;若同構念,擇一或改用其一,並檢查主結果是否因共線而失真。
- [Minor] 維度6(Table 3 表註):星號 *** 未於表註定義顯著水準,補「*** p<0.01」。可直接改。
What ships with it
4 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 12d ago First seen · 98 lines · 500 tokens per session scan A 6fc8329b37fe
thesis-consistency-audit is a skill published in the GitHub repository Nero1688/claude-academic-skills (6 stars, last pushed 9d ago), licensed MIT. It adds 500 tokens to every session and 3,040 once invoked, about $0.0025 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
alterlab-open-notebook
Run Open Notebook, a self-hosted open-source alternative to Google NotebookLM with a full REST API, for AI-powered research and document analysis. Use when organizing research materials into notebooks, ingesting diverse content sources (PDFs, videos, audio, web pages, Office documents), generating AI-powered notes and…
alterlab-treatment-plans
Generates concise (3-4 page), focused medical treatment plans in LaTeX/PDF format across all clinical specialties — general medical treatment, rehabilitation therapy, mental health care, chronic disease management, perioperative care, and pain management — using SMART goal frameworks, evidence-based interventions with…
alterlab-clinical-reports
Writes comprehensive clinical reports — case reports (CARE guidelines), diagnostic reports (radiology, pathology, lab), clinical trial reports (ICH-E3, SAE, CSR), and patient documentation (SOAP notes, H&P, discharge summaries) — with templates, regulatory compliance (HIPAA, FDA, ICH-GCP), and validation tools. Use…
alterlab-pdf-explore
Explore a single PDF in depth — parse it once, then answer questions across its sections, figures, tables, and appendices — comparing methods across sections, extracting every instance of a pattern within the document, and reading values off its charts and tables. Use when interrogating one paper or report end-to-end…
alterlab-markitdown
Convert files and Office documents to clean, LLM-friendly Markdown with Microsoft MarkItDown (markitdown CLI/Python), supporting PDF, DOCX, PPTX, XLSX, images (EXIF + OCR), audio (transcription), HTML, CSV, JSON, XML, ZIP archives, EPUB e-books, and YouTube transcript URLs, with optional AI image descriptions. Use…
alterlab-clinical-decision
Generates professional clinical decision support (CDS) documents for pharmaceutical and clinical research settings — biomarker-stratified patient cohort analyses with outcomes and evidence-based treatment recommendation reports with decision algorithms, supporting GRADE evidence grading, statistical analysis (hazard…