The Agent Skills Directory

[SAFE]: Extensive analysis of the skill's instructions and commands reveals no malicious intent, obfuscation, or unauthorized data access. The behavior is entirely consistent with its stated purpose as a developer utility.
[COMMAND_EXECUTION]: The skill uses the vite-browser CLI tool to perform browser-based diagnostics. This includes subcommands like network to inspect API traffic, screenshot for visual validation, and eval to probe the application's runtime state. These operations are restricted to the local development/debugging environment.
[PROMPT_INJECTION]: The skill was evaluated for indirect prompt injection risks, as it processes external network data.
Ingestion points: External network responses and console logs are ingested via vite-browser network and vite-browser logs (SKILL.md).
Boundary markers: Absent; the agent is not explicitly instructed to ignore instructions within the network responses.
Capability inventory: The agent has the ability to navigate the browser (vite-browser open) and execute JavaScript state probes (vite-browser eval).
Sanitization: No explicit sanitization of the response body is performed before analysis.
Note: This surface is considered a functional requirement for the skill's primary purpose of network debugging and does not escalate the security verdict.

vite-browser-network-regression