← 返回 Skills 市场

Jason Academic Writing

Name: Jason Academic Writing
Author: ithacajason

作者 ITHACAJASON · GitHub ↗ · v1.0.1 · MIT-0

cross-platform ✓ 安全检测通过

107

总下载

当前安装

版本数

在 OpenClaw 中安装

/install jason-academic-writing

功能描述

Complete academic paper writing pipeline with integrity checks and multi-agent review system. Optimized prompts for Methods/Results/Discussion sections. Feat...

使用说明 (SKILL.md)

Academic Writing Pipeline

End-to-end academic paper production with built-in quality gates and multi-agent review.

Pipeline Overview

Research → Write → Integrity Check → Review → Revise → Summary

Each stage has defined inputs/outputs and quality gates. The pipeline is non-linear: stages may loop (Review → Revise → Re-Review) until quality threshold met.

Stage Details

Stage 1: Research

Goal: Gather and organize evidence.

Actions:

Literature search via Semantic Scholar API
Filter by relevance score ≥ 0.5
Grade evidence level (A: meta-analysis, B: RCT, C: observational, D: opinion)
Output: research/evidence.json

Script: scripts/research.py

Stage 2: Write

Goal: Generate structured manuscript.

Actions:

Build argument chain from evidence
Generate sections: Abstract, Introduction, Methods, Results, Discussion
Track citation markers for each claim
Output: draft/manuscript.md

Script: scripts/write.py

Stage 3: Integrity Check (CRITICAL)

Goal: Anti-hallucination verification.

Check types:

Citation verification: DOI exists? Authors match? Year correct?
Data verification: Numbers match tables/figures?
Claim verification: Evidence supports assertion?

Threshold: Must pass 100% of checks to proceed.

Script: scripts/integrity_check.py

APIs used:

Semantic Scholar (https://api.semanticscholar.org)
CrossRef DOI (https://api.crossref.org/works/)

Stage 4: Review (5-Person Panel)

Agents:

Role	Focus	Score Weight
Editor-in-Chief	Contribution, journal fit	30%
Methodology	Methods, stats, reproducibility	25%
Domain Expert	Related work, theory	20%
Devil's Advocate	Strongest counter-arguments	15%
Synthesizer	Merge opinions, roadmap	10%

Decision mapping:

≥80: Accept
65-79: Minor Revision
50-64: Major Revision
\x3C50: Reject

Script: scripts/review.py

Stage 5: Revise

Goal: Address reviewer feedback.

Actions:

Parse Synthesizer roadmap
Generate revision plan with priorities
Rewrite affected sections
Re-run Integrity Check

Script: scripts/revise.py

Stage 6: Process Summary

Goal: Auditable record.

Output: summary.json containing:

Timeline of each stage
Decision points and scores
Integrity check results
Reviewer scores and comments
Revision history

Script: scripts/summary.py

Configuration

Edit config.yaml for:

Model selection (default: qwen3.5-plus)
Temperature (default: 0.3 for stability)
Review thresholds
API keys

Usage

# Full pipeline
python scripts/main.py --topic "your research topic"

# Single stage
python scripts/main.py --stage integrity-check --input draft/manuscript.md

# With custom config
python scripts/main.py --config custom_config.yaml

Quality Gates

Gate	Requirement	Action on Fail
Evidence	≥5 Grade A/B sources	Return to Research
Integrity	100% verification	Return to Write
Review	≥65 score	Loop Revise
Final Integrity	100% verification	Block submission

Key Principles

Integrity First: Citation verification is non-negotiable
Quantified Review: Scores enable objective decisions
Loopable Pipeline: Revision cycles until threshold met
Auditable Output: Process Summary for journal submission

Reference Files

references/review_rubric.md - Detailed scoring criteria
references/evidence_levels.md - Evidence grading standards
references/citation_styles.md - Journal formatting guides

安全使用建议

High-level: the skill appears to do what it says — an end-to-end academic writing pipeline that queries Semantic Scholar/CrossRef and sends manuscript content to an LLM for multi-role review and revisions. Before installing or running: - Review your OPENAI_BASE_URL and OPENAI_API_KEY policies: the pipeline sends full manuscript text to the external LLM; use an API key with appropriate scope and rotate/limit it if needed. Avoid using organization-wide keys if you have privacy concerns. - Check for a .env file in the working directory (or other directories that dotenv might load). The scripts call load_dotenv(), so local secrets in a .env could be read and used; remove or audit .env before running. - Inspect and test in an isolated environment (container or throwaway VM) because the skill writes files and makes network calls. - Be aware the code has brittle JSON-parsing of LLM outputs and at least one partially truncated/buggy function (review.py shows non-robust math and a truncated fallback) — expect runtime errors or unexpected reviewer outputs; plan to validate outputs manually. - If you will process unpublished or sensitive manuscripts, treat outputs and logs carefully: they are stored locally and may be transmitted to third-party APIs. If you want, I can point out the exact lines that call dotenv and where LLM outputs are parsed, or help you harden the scripts (e.g., remove load_dotenv, add explicit JSON-safe wrappers, sanitize/excerpt text sent to LLM).

功能分析

Type: OpenClaw Skill Name: jason-academic-writing Version: 1.0.1 The skill bundle implements a comprehensive academic writing pipeline including research, drafting, citation verification, and a multi-agent review system. It utilizes legitimate external APIs such as Semantic Scholar and CrossRef for data verification and OpenAI for content generation. The code is well-structured and follows safe practices for script orchestration using subprocess argument lists in `scripts/main.py`. While a minor path traversal vulnerability exists in `scripts/main.py` where the user-provided topic is used to create directory names, there is no evidence of malicious intent, data exfiltration, or harmful prompt injection.

能力评估

✓ Purpose & Capability

Name/description (end-to-end academic paper pipeline) align with the included scripts and declared requirements. OpenAI env vars are required because multiple stages call an LLM; research and integrity steps call Semantic Scholar and CrossRef as documented. No unrelated cloud credentials or surprising binaries are requested.

ℹ Instruction Scope

SKILL.md instructs running the included Python scripts and references only pipeline inputs/outputs. Runtime behavior includes sending full manuscript text to an external LLM and calling public APIs (Semantic Scholar, CrossRef) — expected for this purpose. A notable implicit behavior: several scripts call dotenv.load_dotenv(), which will load a .env file from the working directory (and thus can read local secrets not declared in SKILL.md). Also, model output is parsed by regex and then json.loads (fragile) which may misinterpret or leak unstructured text; integrity_check's claim-verification is partly a placeholder. These are scope/robustness issues rather than evidence of misdirection.

✓ Install Mechanism

Instruction-only skill with bundled Python scripts and a requirements.txt. No remote install/downloads, no archive extraction, and no external install URLs. Risk level for installation is low; standard Python dependencies (requests, python-dotenv, openai) are listed.

ℹ Credentials

Declared required env vars (OPENAI_API_KEY, OPENAI_BASE_URL) are appropriate for LLM-driven review/revise stages. Extra behavior: scripts use dotenv.load_dotenv() which may read a .env file and set additional environment variables (e.g., LLM_MODEL). That could result in reading local secrets not documented in the skill metadata. No unrelated credentials (AWS, cloud provider keys, DB passwords) are requested.

✓ Persistence & Privilege

always is false and the skill does not request persistent platform privileges. It writes outputs and reports to a working directory (expected for a pipeline) and does not modify other skill configurations or system-wide settings.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install jason-academic-writing
安装完成后，直接呼叫该 Skill 的名称或使用 /jason-academic-writing 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v1.0.1

- Added requirements.txt to specify required Python packages for installation. - Updated pipeline to require OPENAI_API_KEY and OPENAI_BASE_URL environment variables. - Removed the "Finalize" stage and now generate the process summary as the final step. - Updated SKILL.md to document new requirements and reflect the revised pipeline structure. - Minor fixes and enhancements to research, write, review, and revise script logic.

v1.0.0

- Initial release of end-to-end academic paper writing pipeline with integrity checks and multi-agent review. - Supports structured manuscript generation, automated citation/data/claim verification, and a rigorous multi-role review panel. - Features: self-counterargument framework, overclaim self-audit, bias matrix, and auditable process records. - Configurable for model, review thresholds, and API access; includes clear quality gates and step-by-step usage instructions. - Provides process summary with full audit trail for each manuscript.

元数据

Slug jason-academic-writing

版本 1.0.1

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 2

常见问题