scientific-toolkit-skill

Research computing toolkit for optoelectronic information science and engineering, MATLAB/Octave, Python scientific analysis, signal processing, ima…

执行命令联网读凭据读文件写文件严重 7 · 高危 21zLanqing/codex-claude-academic-skills

它会碰到什么

扫了多少192 个文本文件，2132 KB

它会碰到什么执行命令联网读凭据读文件写文件

命中总数133 处

命中统计严重 7 · 高 21 · 中 38 · 低 24

逐条看命中（28 条严重或高危）

严重 references/scientific-skills/exploratory-data-analysis/references/general_scientific_formats.md:128yaml-unsafe
```
- `yaml`: `yaml.safe_load()` or `yaml.load()`
```
严重 references/scientific-skills/literature-review/SKILL.md:659exec-pipe-to-shell
```
curl -fsSL https://parallel.ai/install.sh | bash
```
严重 references/scientific-skills/networkx/references/io.md:257deserialize-unsafe
```
G = pickle.load(f)
```

严重 references/scientific-skills/paper-lookup/SKILL.md:106cred-paths

2. **Fall back to `.env`** -- check `.env` in the current working directory.

严重 references/scientific-skills/paper-lookup/SKILL.md:106cred-paths

2. **Fall back to `.env`** -- check `.env` in the current working directory.

严重 references/scientific-skills/scikit-learn/references/model_evaluation.md:520deserialize-unsafe
```
loaded_model = pickle.load(f)
```
严重 references/scientific-skills/sympy/references/code-generation-printing.md:424deserialize-unsafe
```
loaded_expr = pickle.load(f)
```
高 references/scientific-skills/citation-management/scripts/extract_metadata.py:32cred-envread
```
self.email = email or os.getenv('NCBI_EMAIL', '')
```
高 references/scientific-skills/citation-management/scripts/extract_metadata.py:162cred-envread
```
api_key = os.getenv('NCBI_API_KEY')
```
高 references/scientific-skills/citation-management/scripts/search_pubmed.py:28cred-envread
```
self.api_key = api_key or os.getenv('NCBI_API_KEY', '')
```
高 references/scientific-skills/citation-management/scripts/search_pubmed.py:29cred-envread
```
self.email = email or os.getenv('NCBI_EMAIL', '')
```
高 references/scientific-skills/literature-review/scripts/generate_pdf.py:46exec-spawn
```
subprocess.run(['pandoc', '--version'], capture_output=True, check=True)
```
高 references/scientific-skills/literature-review/scripts/generate_pdf.py:91exec-spawn
```
result = subprocess.run(cmd, capture_output=True, text=True, check=True)
```
高 references/scientific-skills/literature-review/scripts/generate_pdf.py:110exec-spawn
```
subprocess.run(cmd.split(), capture_output=True, check=True)
```
高 references/scientific-skills/pymatgen/scripts/phase_diagram_generator.py:39cred-envread
```
api_key = os.environ.get("MP_API_KEY")
```
高 references/scientific-skills/timesfm-forecasting/scripts/check_system.py:138exec-spawn
```
result = subprocess.run(
```
高 references/scientific-skills/timesfm-forecasting/scripts/check_system.py:186exec-spawn
```
result = subprocess.run(
```
高 references/scientific-skills/timesfm-forecasting/scripts/check_system.py:307cred-envread
```
hf_cache = os.environ.get("HF_HOME", os.path.expanduser("~/.cache/huggingface"))
```
高 references/scientific-skills/xlsx/scripts/office/soffice.py:14exec-spawn
```
subprocess.run(["soffice", ...], env=env)
```
高 references/scientific-skills/xlsx/scripts/office/soffice.py:26cred-envread
```
env["SAL_USE_VCLPLUGIN"] = "svp"
```
高 references/scientific-skills/xlsx/scripts/office/soffice.py:30cred-envread
```
env["LD_PRELOAD"] = str(shim)
```
高 references/scientific-skills/xlsx/scripts/office/soffice.py:37exec-spawn
```
return subprocess.run(["soffice"] + args, env=env, **kwargs)
```
高 references/scientific-skills/xlsx/scripts/office/soffice.py:59exec-spawn
```
subprocess.run(
```
高 references/scientific-skills/xlsx/scripts/office/validators/redlining.py:138exec-spawn
```
result = subprocess.run(
```
高 references/scientific-skills/xlsx/scripts/office/validators/redlining.py:167exec-spawn
```
result = subprocess.run(
```
高 references/scientific-skills/xlsx/scripts/recalc.py:34exec-spawn
```
subprocess.run(
```
高 references/scientific-skills/xlsx/scripts/recalc.py:55exec-spawn
```
subprocess.run(
```

高 references/scientific-skills/xlsx/scripts/recalc.py:92exec-spawn

result = subprocess.run(cmd, capture_output=True, text=True, env=get_soffice_env())

这一栏是扫描器报的事实，不是结论。命中多不等于有毒（安全工具、规则库、示例脚本本来就会包含危险写法），命中少也不等于干净。它和你手上的凭据、文件、网络有什么关系，需要你自己看。

技能内容

Scientific Toolkit Skill

Scope

Use this skill for科研计算 and software-assisted research:

MATLAB/Octave scripts, debugging, refactoring, signal/image processing, FFT, filtering, matrix computation, simulation, and figure export.
Python scientific workflows with NumPy, SciPy, pandas, matplotlib, seaborn, scikit-learn, statsmodels, SymPy, and related tools.
Statistics, exploratory data analysis, sensor/time-series forecasting, optimization, discrete-event simulation, quantum optics/open quantum systems, materials data, and graph/network analysis.
Literature lookup, citation metadata, BibTeX, and reference verification when it supports coding or research analysis.

Use research-writing-skill for manuscript prose. Use office-academic-skill for Word/PPT deliverables.

Domain Defaults

The user's field is光电信息科学与工程. Prefer examples and checks relevant to:

Optics, optoelectronics, optical communication, optical sensing, fiber sensing, BOTDR/BOTDA, BGS, SPM, dispersion, noise, and deconvolution.
Signal processing, image processing, spectroscopy, detector data, sensor time series, calibration, and uncertainty.
MATLAB simulation and reproducible figure generation for论文/答辩.

Do not fabricate physical parameters, material constants, software menu operations, experimental results, or paper conclusions. When uncertain, ask for the source file or mark the assumption.

General Workflow

Read the provided code, data, README, docs, and project instructions before changing anything.
Identify variables, dimensions, units, input/output paths, random seeds, and expected figures.
Make small, verifiable changes and avoid unrelated refactors.
Prefer mature libraries over hand-rolled numerical methods.
Run a script-level or test-level verification when possible.
Report environment, commands, output paths, generated figures, and known limitations.

MATLAB And Figures

Preserve the original code structure when possible.
Add concise comments for physical meaning, units, assumptions, or formula sources.
Centralize key parameters and avoid hardcoded absolute paths.
Add rng for stochastic simulations when reproducibility matters.
For publication figures, export both high-resolution .png and vector .svg when feasible.
Check axes, units, legends, sampling rate, line width, font, color, and image resolution.

For MATLAB/Octave details, use references/scientific-skills/matlab/SKILL.md.

Python Scientific Modules

Load only the relevant bundled reference:

Plotting and publication figures: matplotlib, seaborn, scientific-visualization.
Statistics and time series: statistical-analysis, statsmodels, timesfm-forecasting.
Machine learning: scikit-learn.
Symbolic math and formulas: sympy.
Exploratory data analysis: exploratory-data-analysis.
Optimization: pymoo.
Simulation: simpy.
Quantum optics/open quantum systems: qutip.
Materials/crystal/band/DOS workflows: pymatgen.
Graphs/networks/citation graphs: networkx.
FITS or astronomical/optical imaging style data: astropy.
Spreadsheet/PDF utilities: xlsx, pdf.
Literature/citation support: paper-lookup, citation-management, literature-review.

Some bundled references mention optional installs such as uv pip install ... or optional API keys for higher rate limits. Do not install packages, use cloud APIs, or send user data to external services unless the current task requires it and the user agrees.

Safety Rules

Never expose or commit API keys, tokens, private data, or unpublished paper content.
Do not overwrite original data, code, Word/PPT, or figures. Write versioned outputs.
Do not delete or recursively clean user files without explicit confirmation.
For external lookups, prefer open APIs and official documentation; clearly distinguish live lookup results from local inference.

Verification

For code:

Run the relevant script or a minimal example.
Check generated files exist and are readable.
Inspect plots for axes, units, legends, and plausible dimensions.

For research analysis:

State software versions when known.
List input files and commands.
Mark assumptions and uncertain parameters.

想直接用这个技能？

本站把开放许可（MIT / Apache 等）的技能按仓库打包整理到网盘，点一下转存到你自己的网盘，不用一个个从 GitHub 拉。许可未声明的技能只给原始仓库链接，不打包。

去原始仓库

它属于哪个仓库

仓库zLanqing/codex-claude-academic-skills

星标★ 3,961

本站分层T1

该仓技能数24

原文件路径scientific-toolkit-skill/SKILL.md

同一个仓库里的其他技能

看这个仓库的全部 24 个技能