First watch condition: Whether the project documents reproducible sandbox settings and validation showing what network or filesystem access the language servers actually have. The configuration is a narrow editor project, not evidence that the broader agent security problem is solved.
Thesis
The nv configuration treats the editor as a review surface for AI-authored code, pairing visible changes with constrained language-server operation.
Observed facts
The repository describes nv as a Neovim 0.12 configuration for a workflow where Claude Code writes code and a person reads, checks, and edits it.
1The repository says all 11 language servers run in a sandbox without network access.
1The repository says completion suggestions are not inserted until the user presses Enter or Tab.
1
Mechanism
The editor combines sandboxed language services with explicit user confirmation and visibility into external file changes, making inspection part of the workflow rather than relying solely on the coding assistant's behavior.
Why now
AI-assisted coding increases the importance of review interfaces and local boundaries around supporting tools.
Implications
A coding setup can make human verification easier by exposing edits and requiring deliberate acceptance of completions.
Evidence boundary
These are repository descriptions; they do not establish independent security testing, effectiveness, or user adoption.
Watch conditions
Whether the project documents reproducible sandbox settings and validation showing what network or filesystem access the language servers actually have.