Borrowing it
Nothing to install: this file belongs to u9401066/med-paper-assistant. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/u9401066/med-paper-assistant/master/.claude/skills/auto-paper/SKILL.mdgit clone --depth 1 https://github.com/u9401066/med-paper-assistantWrote 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/u9401066/med-paper-assistant/auto-paper)<a href="https://agentmods.dev/skills/u9401066/med-paper-assistant/auto-paper"><img src="https://agentmods.dev/badge/skills/u9401066/med-paper-assistant/auto-paper/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/u9401066/med-paper-assistant/auto-paper"><img src="https://agentmods.dev/badge/skills/u9401066/med-paper-assistant/auto-paper.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 6 findings, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Prompt Injection · line 60 Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.Fix: Remove the large whitespace padding (blank-line blocks or long space runs) and review any content hidden below or to the right of it. Keep skill files compact and reviewable so no instructions can be
- medium Tool Misuse · line 364 Tool defaults are unsafe or overly permissive (e.g. disabled TLS verification, no authentication, world-writable permissions). Unsafe defaults widen the attack surface.Fix: Override unsafe defaults with secure settings (verify=True, auth required, restrictive permissions). Review and harden all tool configurations.
- medium Agent Snooping · line 1985 Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.Fix: Remove all code or instructions that list or read other skills' files or directories. Skills should operate independently; cross-skill access is a privilege escalation.
- medium Agent Snooping · line 1989 Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.Fix: Remove all code or instructions that list or read other skills' files or directories. Skills should operate independently; cross-skill access is a privilege escalation.
- medium Prompt Injection · line 2122 Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.Fix: Remove the large whitespace padding (blank-line blocks or long space runs) and review any content hidden below or to the right of it. Keep skill files compact and reviewable so no instructions can be
- medium Prompt Injection · line 2133 Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.Fix: Remove the large whitespace padding (blank-line blocks or long space runs) and review any content hidden below or to the right of it. Keep skill files compact and reviewable so no instructions can be
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 | $0.00089 | $0.34632 |
| Opus 5 | $0.00044 | $0.17316 |
| Sonnet 5 | $0.00018 | $0.06926 |
| Haiku 4.5 | $0.00009 | $0.03463 |
Grade A, and why
auto-paper 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 7d 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.
How it starts
The opening of the file, as written. The whole thing — 2,219 lines — stays where its author put it; the contents beside it link to each section on GitHub.
全自動論文撰寫 + 閉環自我改進
閉環架構
Instructions (AGENTS.md) → Skill (auto-paper) → Writing (drafts/) → Hooks (audit) → 回頭更新 Skill/Instructions
核心理念（CONSTITUTION §22）：可審計、可拆解、可重組。詳見 AGENTS.md。
四個審計維度：
| 維度 | 檢查 | 更新對象 |
|---|---|---|
| Paper | 引用、字數、Anti-AI | draft_action(action="patch") 修正論文 |
| Methodology | 方法學可再現性 | 補充方法學細節 |
| Compliance | Agent 是否遵循流程 | .memory/ 記錄偏差 |
| Meta | Skill/Hook/Instruction | 更新 SKILL/Hook/AGENTS |
執行審計軌跡（Audit Trail）
每次執行在 projects/{slug}/.audit/ 產出：
| 檔案 | 內容 |
|---|---|
pipeline-run-{ts}.md |
Phase 摘要 + Hook 統計 + Decision Log |
search-strategy.md |
搜尋策略和結果 |
reference-selection.md |
文獻篩選決策 |
concept-validation.md |
Novelty 驗證過程 |
hook-effectiveness.md |
Hook 觸發率/通過率/誤報率 |
quality-scorecard.md |
方法學 + 文字品質（0-10 分） |
checkpoint.json |
斷點恢復：last_completed_phase, phase_outputs |
恢復邏輯：偵測 checkpoint.json → 詢問從 Phase N+1 繼續或重來。
雙重 Hook 系統
Copilot Hooks（A-D，寫作時即時修正）定義於本檔。
Pre-Commit Hooks（P1-P8 + G1-G7）定義於 git-precommit/SKILL.md。
兩者互補：Copilot 處理細節，Pre-Commit 處理全局。
🚧 Hard Gate Enforcement（Code-Level，不可跳過）
SKILL.md 是 soft constraint（Agent 可能忽略）。以下 MCP Tools 是 code-enforced hard limits。
必要 MCP Tool 呼叫
| 時機 | MCP Tool | 說明 |
|---|---|---|
| 每個 Phase 完成後 | pipeline_action(action="validate_phase", phase=...) |
返回 PASS/FAIL + 缺少的 artifact，FAIL 則禁止進入下一 Phase |
| Phase 7 每輪開始 | pipeline_action(action="start_review") |
啟動 AutonomousAuditLoop 狀態機，返回 round context |
| Phase 7 每輪結束 | pipeline_action(action="submit_review", scores=...) |
提交分數，返回 verdict (CONTINUE/QUALITY_MET/MAX_ROUNDS/REWRITE_NEEDED) |
| Phase 7 需要大幅重寫 | pipeline_action(action="rewrite_section", sections=..., reason=...) |
回退到 Phase 5 重寫指定 section，最多 2 次回退 |
| Phase 5 每 section 完成後 | pipeline_action(action="approve_section", section=..., decision=...) |
Autopilot: Agent 自我審閱後自動 approve。手動: 用戶審閱 → approve 或 revise。Phase 5 gate 要求全部 approved |
| 用戶要求暫停時 | pipeline_action(action="pause", reason=...) |
暫停 pipeline，記錄 draft hash 以偵測用戶編輯 |
| 暫停後恢復 | pipeline_action(action="resume") |
恢復 pipeline，偵測用戶修改並建議重新驗證 |
| Pipeline 中途任意時刻 | pipeline_action(action="heartbeat") |
返回全 Phase 狀態 + 剩餘工作項，Agent 無法自稱 "done" |
| Agent 需要一次總檢查時 | pipeline_action(action="doctor") |
返回 11-phase readiness、外部 MCP 宣告/command 可用性、最近 gate cache |
| Phase 5 每次 Hook 評估後 | record_hook_event(hook_id, event_type) |
記錄 A/B/C/E Hook 的 trigger/pass/fix/false_positive，Phase 6 gate 會驗證有實際記錄 |
| Phase 6 之前（審計階段） | run_quality_checks(action="quality_audit", scores=...) |
設定 ≥4 維度品質分數 + 產生 scorecard/hook-effectiveness 報告，Phase 6 gate 驗證分數數據 |
| Phase 10 之前（自我改進） | run_quality_checks(action="meta_learning") |
執行 D1-D9 分析 + 寫入 meta-learning-audit.yaml / meta-learning-feedback.yaml，Phase 10 gate 驗證 analysis_steps |
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.
- 7d ago Changed · -4 lines 3c5bb370b790
- 11d ago First seen · 2,223 lines · 89 tokens per session scan A 9f968081602b
auto-paper is a skill published in the GitHub repository u9401066/med-paper-assistant (12 stars, last pushed 9d ago), licensed Apache-2.0. It adds 89 tokens to every session and 34,632 once invoked, about $0.0004 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 2026-08-30.
Other skills, from other repositories
citation-manager
Format and manage citations — APA, MLA, Chicago, IEEE, and Vancouver styles with BibTeX export.
manuscript-reframe
Reframe report-like academic drafts into paper-form scientific arguments while preserving or explicitly renegotiating author intent; requires an approved old-versus-proposed spine, evidence and argument baselines, analysis-role control, and post-edit drift review.
peer-review
Review another author's manuscript, paper, thesis chapter, proposal, or preprint as an external reviewer. Use when asked to evaluate novelty, significance, gap-contribution fit, claim-evidence adequacy, methods, evaluation, overclaim risks, structure, writing, required revisions, or recommendation without rewriting…
revision-escalation
A revision-control guide for writing, coding, manuscripts, rebuttals, and restructuring work. It says to stop applying local edits after the same problem remains unresolved through three attempts and check the underlying requirements or structure.
verify
Fact-check claims encountered during reading — dates, names, events, citations. Use when encountering historical facts or disputed claims.
author
Scaffold a conforming AI-layer artifact — a rule, skill, hook, agent, or output style — that already passes the format checks, then verify it. Use when adding any new AI-layer artifact — a shared one under .agents/ or a Claude-only one under .claude/ — so it matches the existing shape instead of drifting.