academic-pipeline-v1
Security checks across static analysis, malware telemetry, and agentic risk
Overview
This is an instruction-only academic workflow orchestrator with no code or credentials, but it includes a de-AI humanizing stage and sends manuscript content through searches and other skills, so users should review it carefully before use.
Before installing, decide whether the humanize/de-AI stage complies with your academic or publication rules, and review the dependent skills separately. Do not run the WebSearch-based originality or claim checks on confidential drafts unless you are comfortable sharing excerpts with external services, and keep generated process records secure.
Publisher note
Academic Pipeline v3.5 全流程学术研究与论文写作编排器 12 阶段全景图： Stage 1 LITERATURE SEARCH (IMA + academic-search) │ │ ↓ 用户确认 │ │ Stage 2 RESEARCH (deep-research) │ │ ↓ 用户确认 │ │ Stage 3 WRITE (academic-paper) │ │ ↓ 用户确认 │ │ Stage 4 INTEGRITY ✦ 强制 ✦ (引用/数据 100% 验证) │ │ ↓ PASS → 继续；FAIL → 修复重验（最多 3 轮） │ │ Stage 5 REVIEW (5 人同行评审) ✦ 强制 ✦ │ │ ↓ Accept→Stage 9 / Minor→Stage 6 / Reject→Stage 3 │ │ Stage 6 REVISE (academic-paper revision 模式) │ │ ↓ 用户确认 │ │ Stage 7 RE-REVIEW (验证评审) ✦ 强制 ✦ │ │ ↓ Accept→Stage 9 / Major→Stage 8 │ │ Stage 8 RE-REVISE (二次修订，最多 1 轮) │ │ ↓ 用户确认 │ │ Stage 9 FINAL INTEGRITY ✦ 强制 ✦ (零问题才能过) │ │ ↓ PASS │ │ Stage 10 HUMANIZE ✦ 强制 ✦ (去 AI 化处理) │ │ ↓ 用户确认去 AI 版本 │ │ Stage 11 FINALIZE (academic-paper format-convert) │ │ ↓ MD+DOCX→LaTeX→PDF │ │ Stage 12 PROCESS SUMMARY (过程记录 PDF)
Static analysis
No static analysis findings were reported for this release.
VirusTotal
VirusTotal findings are pending for this skill version.
Risk analysis
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
A user could unintentionally violate academic or publication rules, or mislead reviewers, by using the humanizing step as more than ordinary style editing.
The pipeline explicitly includes a full-paper de-AI/humanizing stage. In an academic submission workflow, this can be used to make AI-assisted writing appear non-AI-assisted, and the provided artifacts do not show a guardrail requiring disclosure or checking the user's journal/institution policy before this step.
**10** | **HUMANIZE** | **`humanizer` + `humanizer-zh`** ... **De-AI 化论文全文 + 双语变更摘要**
Treat this stage only as transparent style editing, disclose AI assistance where required, and skip it if the target institution or venue prohibits AI text humanization.
Confidential research text, distinctive phrases, author names, or unpublished claims may be exposed to external search services.
The originality check sends selected manuscript sentences or fragments to WebSearch. This is purpose-aligned for plagiarism screening, but it means draft or unpublished text may leave the local conversation/tool boundary.
Use the characteristic sentence (or key fragment) as a WebSearch query ... Review the top 5-10 search results
Use this only when you are comfortable sending excerpts to external search providers; redact sensitive content or ask the agent to confirm before each external search.
Private drafting decisions, reviewer comments, and evidence of AI assistance may be preserved in generated records.
The process summary stores detailed collaboration history and user quotes. This is useful for reproducibility, but it can contain sensitive research context and may influence or be reused in later workflow stages.
generates structured MD summarizing full human-AI collaboration history with user quotes, key decisions, iteration details
Store generated process records securely, review them before sharing, and delete or redact sensitive details when they are no longer needed.
Downstream skills may read drafts, call APIs, write files, or have their own security/privacy behavior outside this review.
The orchestrator delegates substantive work to multiple external skills. That is central to its purpose, but the permissions and behavior of those dependencies are not established by this artifact set.
depends_on: "ima-skills, academic-search, deep-research, academic-paper, academic-paper-reviewer, humanizer, humanizer-zh"
Review and install only trusted versions of the dependent skills before running the full pipeline.