The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill manages BrowserStack credentials by instructing the user to provide BROWSERSTACK_USERNAME and BROWSERSTACK_ACCESS_KEY via environment variables. These are used to connect to BrowserStack's official domain (cdp.browserstack.com). Using environment variables instead of hardcoded secrets is a standard security practice.
[REMOTE_CODE_EXECUTION]: The skill installs the official browserstack-local package from the npm registry to enable local testing capabilities. As this is a well-known package from a trusted provider, it is considered a routine dependency.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data from the BrowserStack API, creating a potential surface for indirect prompt injection.
Ingestion points: Content is retrieved from the BrowserStack API using MCP tools like browserstack_get_logs and browserstack_get_sessions (SKILL.md).
Boundary markers: The instructions do not define specific delimiters to wrap or isolate this external content.
Capability inventory: The agent has the ability to write to the local filesystem (playwright.config.ts) and execute shell commands (npx playwright test).
Sanitization: There is no mention of sanitizing or validating the ingested log data before it is processed by the agent.

browserstack