python-scripts

name	python-scripts
description	Write and execute Python scripts in this workspace, installing any packages they need with conda or pip. Use for data analysis, charts, quick computations, scraping/parsing research sources, validating numbers cited in the book, or any one-off automation that benefits from running real Python.
argument-hint	Describe what the script should do, any input/output files, and which libraries (if known). Mention whether you want a reusable script saved or a one-off run.
user-invocable	true

Python Scripts Skill

Author, install dependencies for, and run Python scripts inside the aidou workspace.

When to Use

Use this skill when a task benefits from executing real Python, for example:

• Verifying or computing numbers/statistics cited in the book before writing them.
• Parsing, cleaning, or summarising research sources (Markdown, HTML, PDF, CSV, JSON).
• Generating charts or tables from data.
• Quick automation: file processing, bulk renames, link checking, scraping.
• Prototyping or sanity-checking an algorithm described in the text.

Trigger phrases: "run a python script", "compute", "plot", "parse this data", "check these numbers", "scrape", "generate a chart".

Environment (verified 2026-06-30)

Item	Value
Package manager	conda 26.3.2 (Miniconda at C:\Users\ctham\AppData\Local\miniconda3)
Active env	base (the only environment)
Python	3.13.13
pip	26.0.1
conda channels	conda-forge, defaults
Platform	Windows, win-64, PowerShell

Both pip install and conda install work. PyPI and the conda channels are reachable.

[!IMPORTANT] TLS verification is disabled on this machine (corporate proxy / self-signed certificate). conda commands print InsecureRequestWarning for conda.anaconda.org and repo.anaconda.com. Installs still succeed, but over unverified HTTPS. This is expected here — do not treat the warning as a failure.

[!IMPORTANT] PowerShell exit-code quirk. conda writes warnings to stderr, which makes PowerShell raise a RemoteException and return exit code 1 even when the command succeeded. Never judge success by exit code alone — inspect stdout (the package list, "Would install…", or solved-environment block) to confirm.

Workflow

1. Decide scope. A one-off computation can run inline with python -c "…". Anything reusable or longer than a few lines: save a .py file (see layout below).
2. Check for the package first. Before installing, run conda list <pkg> or python -c "import <pkg>" — it may already be present.
3. Install only what's missing. Prefer conda install -c conda-forge <pkg> for scientific/binary packages (numpy, pandas, matplotlib, scipy, lxml). Use pip install <pkg> for pure-Python or PyPI-only packages.
4. Write the script, then run it and inspect stdout.
5. Clean up temporary scratch scripts unless the user wants them kept.

Installing packages

Check first, then install only if missing:

conda list pandas                       # already installed?
conda install -c conda-forge pandas     # binary/scientific deps -> conda
pip install some-pure-python-pkg        # PyPI-only -> pip

To confirm an install actually worked despite the exit-code quirk, verify by import:

python -c "import pandas, sys; print(pandas.__version__)"

If conda's stderr noise makes terminal output hard to read, filter the warnings:

conda install -c conda-forge pandas 2>&1 |
  Select-String -Pattern "InsecureRequestWarning|warnings.warn" -NotMatch

For an unattended install, add -y to skip the conda confirmation prompt.

Running scripts

Inline one-off:

python -c "print(sum(range(101)))"

Saved script:

python scripts/<name>.py

Conventions for this repo

• Save reusable scripts in scripts/ (e.g. scripts/check_numbers.py), matching the existing scripts/html2md.ps1.
• Read inputs from research/, sources/, or book/ with relative paths so scripts are portable.
• Keep scripts small and single-purpose; print results to stdout.
• Don't add a requirements.txt or new env unless the user asks — install into base.
• Never commit large generated data/binaries without asking.

Worked example

Count words in a chapter:

# scripts/wordcount.py
from pathlib import Path

text = Path("book/02-productivity.md").read_text(encoding="utf-8")
print(f"{len(text.split()):,} words")

python scripts/wordcount.py

Troubleshooting

Symptom	Cause	Fix
Exit code 1 but output looks fine	PowerShell + conda stderr warnings	Ignore exit code; check stdout / verify by import
InsecureRequestWarning flood	TLS verification disabled (proxy)	Expected; filter with Select-String … -NotMatch
Solving environment: failed for a real package	Package not on configured channels	Try -c conda-forge, or fall back to pip install
ModuleNotFoundError after install	Installed into a different interpreter	Run python -c "import sys; print(sys.executable)" to confirm it's the Miniconda base python