pythonista-reviewing_skill
- Shell
1
GitHub Stars
1
Bundled Files
2 months ago
Catalog Refreshed
4 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill gigaverse-app/skillet --skill pythonista-reviewing- SKILL.md4.2 KB
Overview
This skill helps you perform focused, high-value code reviews on PRs and diffs. It prioritizes structural, cross-file, and egregious issues over nitpicky style comments. The goal is fast, actionable feedback that surfaces real technical risks and candidate fixes.
How this skill works
The skill inspects aggregated diffs from the branch vs main and looks for cross-file patterns such as duplicated logic, god functions, weak typing patterns, and dead code. It filters out noisy files (lock files, docs) and emphasizes items that require human judgment. For large PRs it recommends or drives large-context LLM review runs and produces a numbered list of suggested actions for human approval.
When to use it
- When reviewing pull requests or diffs larger than a few files
- When you want cross-file pattern detection (duplication, architectural issues)
- Before merging to catch egregious problems (god functions, dead code, weak types)
- When tests look suspiciously shallow or only assert trivial behavior
- When preparing a PR for an automated or LLM-assisted review pass
Best practices
- Review at the PR level, not file-by-file; focus on structural problems
- Filter out lock files and docs to reduce noise before analysis
- Flag duplication and repeated patterns first; they indicate refactor opportunities
- Prioritize cross-file, architectural issues and avoid nitpicks (naming, whitespace)
- Use LLMs as an assistant: treat suggestions as hints and validate before applying
- Always present findings with numbered suggested actions and ask for approval before making changes
Example use cases
- Run a full-diff review to detect repeated validation logic spread across modules
- Identify a god function over 100 lines that should be split for testability
- Spot weak typing patterns (raw dicts, hasattr) that hide bugs and suggest type annotations
- Use a large-context LLM to surface cross-file bloat or AI-generated repetitive code
- Produce a concise action list after review for the author to approve and implement
FAQ
No — avoid these minor style issues unless they hide real confusion; focus on structural and functional risks.
How should LLM findings be used?
Treat them as hints: validate with a second model or human review and then convert findings into numbered action items for approval.
