Feedback: commands flagged incorrectly (false positives)

Report specific commands that asked for approval when they shouldn't have.

In the current release, part of the permission check uses a model to judge whether a command is risky, and it errs on the side of caution — so it can flag things that are plainly harmless. Those reports are exactly what we want here.

```
Command:
Prompt shown:
Why it's safe:
Version / OS / shell:
```

One command per comment. Redact anything sensitive.

Note: some prompts are still intentional — a command can be safe itself but still reach the network, credentials, or files outside the workspace.

For broader permission problems, use #2542.

Utilisez les sections Command, Prompt shown, Why it's safe et le modèle Version / OS / shell de l'issue pour documenter une demande d'approbation considérée à tort comme valide. Occultez les informations sensibles et publiez une commande par commentaire ; les problèmes d'autorisation plus généraux doivent être traités dans #2542.

Rédigé par le modèle d'indexation à partir du texte de l'issue.