The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes VS Code launch scripts (scripts/code.sh or scripts/code.bat) and npx @playwright/cli to drive automation. These are standard development tools and are constrained to a performance-testing context.
[EXTERNAL_DOWNLOADS]: The skill uses Playwright-core and standard Node.js libraries. It mentions npx @playwright/cli for inspection. These are official development tools from Microsoft and are well-known services.
[DATA_EXFILTRATION]: The skill manages VS Code user data profiles, which may contain authentication secrets (e.g., Copilot auth). However, the skill provides explicit warnings to keep these folders inside gitignored .build/ directories and never attach them to PRs. It does not contain code to send this data to external untrusted domains.
[PROMPT_INJECTION]: The skill provides instructions to the agent on how to interact with the LLM via the automated VS Code Chat interface. These instructions include safety warnings about using read-only commands and throwaway workspaces, reinforcing rather than bypassing safety protocols.
[CREDENTIALS_UNSAFE]: While the skill manages authentication profiles for testing purposes, it does not include hardcoded secrets or instruct the user to provide them unsafely; it leverages existing local VS Code authentication states.

auto-perf-optimize