Encuentra los errores
Guess first, no penalty. You asked an AI to add a row to a pandas table and retry an upload. Your project pins pandas 2.2. Tap the two lines to check before you use them.
The AI's suggestion
0 de 2 encontrados
A package is code someone else published for you to install from a registry, the official index of packages, such as npm for JavaScript or PyPI for Python. AI tools sometimes invent package names that sound right. Attackers know this, so they can register those names and fill them with harmful code.
AI tools also learned from old code, so they suggest functions that were removed years ago. Your lockfile (the file that pins the exact versions your project uses, such as package-lock.json or poetry.lock) tells you which version's docs apply.
Before you install or trust anything new:
- Find it on the registry. Check the maintainer, the source repository and how widely it's used.
- Check the version. Read the docs for the version in your lockfile, not the latest one.
- Prefer what you have. If the standard library or an existing dependency does the job, use that.
A worked example
The notes one developer wrote before using the suggestion above:
upload-retry-helper: not something we use. Not installing. A loop with
a timeout covers the retry in five lines.
df.append: lockfile says pandas 2.2. The 2.0 release notes say append
was removed, use concat. Changed to:
df = pd.concat([df, pd.DataFrame([new_row])], ignore_index=True)Completa los espacios
Fill the gaps.
Checklist
Before I install a package an AI suggested, I find it on the and check its . I read the docs for the version in my .
Try it
Tus pasos, unos 1 minutos
- 1.Check every package and function an AI suggested in your last task against the registry and your lockfile. Add the checklist above to your kit.1 min
Your AI code review kit is now complete. Keep it next to your editor:
1. Ask for tests: normal, empty, bad and boundary cases, exact values.
2. Ask for code: task, context, rules, then code + edge cases + tests.
3. Diff review: scope, errors, edge cases, your codebase, security.
4. Logic check: trace one small, mixed example by hand. Keep it as a test.
5. Package check: registry, maintainer, lockfile version, use what you have.