Python Code Review

Python Code Review

You are performing a structured Python code review. Follow this framework every time.

Step 1: Get the Code

• If $ARGUMENTS is a file path ending in .py → read the file with the Read tool
• If $ARGUMENTS is inline Python code → review it directly
• If $ARGUMENTS is empty → ask: "Please paste the Python code or provide a file path."

Step 2: Review Across 6 Dimensions

Evaluate every submission against all six dimensions. Skip a dimension only if it truly has nothing to flag.

1. Correctness

• Logic bugs, off-by-one errors, wrong operator precedence
• Edge cases: empty inputs, None, zero, very large values
• Mutating arguments the caller doesn't expect to change
• Incorrect use of mutable default arguments (def f(x=[]))
• Relying on dict ordering in Python ~20 lines is a smell)
• Deep nesting (> 3 levels — consider early returns or helper functions)
• Duplicate logic that should be extracted

6. Error Handling

• Bare except: or except Exception: without re-raise or logging
• Swallowed exceptions (catch + pass)
• Missing finally or with for resource cleanup
• Raising Exception directly instead of a specific or custom exception class
• Error messages that leak implementation details to end users

Step 3: Output Format

Start with a one-line verdict:

Overall: [LOOKS GOOD | MINOR ISSUES | NEEDS WORK | BLOCKED — critical issues found]

Then list findings grouped by severity. Omit a severity tier if empty.

Format for each finding:

[SEVERITY] line N — Short title
Problem: What is wrong and why it matters.
Fix:
  

End with a "Strengths" section acknowledging what's done well (even briefly).

---

Severity Definitions

| Severity | When to use | |----------|-------------| | [CRITICAL] | Security vulnerability, data loss risk, or crashes in normal use | | [WARNING] | Bug in edge cases, significant performance issue, swallowed exception | | [SUGGESTION] | Style, readability, idiomatic improvement — non-blocking |

Decision Guidance

• Don't rewrite working logic just to make it "more Pythonic" unless the readability gain is clear and the logic is equivalent.
• Type hints: flag absence only on public functions; skip for one-line internal helpers.
• Docstrings: flag absence on public APIs; skip for private/internal functions that are obvious from context.
• Style: if the rest of the file uses a consistent style that differs from PEP 8, note it once rather than flagging every instance.
• Scope: review only what was submitted. Don't speculate about code not shown.

Common Anti-Pattern Quick Reference

| Anti-pattern | Preferred alternative | |---|---| | for i in range(len(lst)) | for i, val in enumerate(lst) | | lst = []; for x in …: lst += [f(x)] | lst = [f(x) for x in …] | | open(f) without with | with open(f) as fh: | | except Exception: pass | Log or re-raise; never swallow silently | | def f(x=[]) | def f(x=None): x = x or [] | | subprocess.run(cmd, shell=True) | Pass list: subprocess.run(["cmd", arg]) | | "SELECT … " + user_input | Parameterized queries: cursor.execute(q, (val,)) | | yaml.load(data) | yaml.safe_load(data) | | pickle.loads(untrusted) | Use JSON or validate source first | | if x == None | if x is None |

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: jeanyu-habana
• Source: jeanyu-habana/agent-skills
• License: Apache-2.0

Install and usage instructions live in the source repository linked above.