python-reviewer — Claude Code subagent: features and setup
python-reviewer reviews Python visual scripts before placing them into Power BI. It catches banned libraries, layout clipping, and missing display calls.
GPL-3.0 ★ 1,104 updated 2026-10-28
.github/agents/python-reviewer.agent.mdWhat python-reviewer does
Running on the Sonnet model, this subagent relies on Read, Grep, and Glob to inspect Python visualization files. Its primary goal is ensuring code aligns with Power BI constraints before you test it on a canvas. Network requests or file downloads cause immediate failures because the target runtime environment forbids outbound traffic.
The review enforces strict platform boundaries. Scripts must not instantiate a dataset variable, as Power BI injects this DataFrame automatically. Only approved packages pass: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, and pillow. Plotly, Bokeh, and Altair are flagged as errors.
Formatting checks require proper 72 DPI dimensions, muted colors, and a tight_layout() call to prevent clipped labels. Omitting plt.show() on the final line leaves the visual completely blank, so the agent marks the script as needing changes until that call is added.
What it helps with
- validate Power BI Python visual scripts before testing them on a canvas
- block unauthorized packages and network calls in visualization code
- prevent chart clipping by checking 72 DPI dimensions and tight layout calls
- catch missing plt.show calls and unhandled empty datasets
Example prompts for the agent
- Review my distribution plot script with python-reviewer before putting it into Power BI
- Run python-reviewer on
visual.pyto check for Power BI compatibility
How to install python-reviewer
Manually: put the files in place
.github/agents/python-reviewer.agent.md1Plugin marketplace
/plugin marketplace add data-goblin/power-bi-agentic-development/plugin install reports@power-bi-agentic-development2Manually: put the files in place
.claude/agents/python-reviewer.mdFor all projects use ~/.claude/agents/. Claude delegates by the description field or an @-mention. Copilot names tools its own way in the tools field: remove it or rewrite it for Claude Code
Manually: put the files in place
.codex/agents/python-reviewer.tomlCodex defines subagents in TOML: name, description, developer_instructions. Move the Markdown text into developer_instructions by hand. For all projects use ~/.codex/agents/
Manually: put the files in place
.cursor/agents/python-reviewer.mdCursor also reads .claude/agents/
Step by step, with commands for each platform: Claude Code subagents: how to create one, call it and keep context clean
Questions about python-reviewer
How to install python-reviewer in Claude Code?
Run /plugin marketplace add data-goblin/power-bi-agentic-development, then run /plugin install reports@power-bi-agentic-development.
What does python-reviewer do?
It reviews matplotlib and seaborn scripts against Power BI runtime rules, flags invalid imports or missing display calls, and returns a READY or NEEDS CHANGES verdict.
Which Python libraries are allowed by python-reviewer?
The subagent allows matplotlib, seaborn, numpy, pandas, scipy, scikit-learn, statsmodels, and pillow. Interactive plotting engines such as Plotly or Bokeh fail the review.