The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various shell commands using npx, npm, and bash. Specifically, it runs npx @playwright/mcp@latest for interactive browser automation and npx playwright test for headless testing. It also uses bash to interact with the Mail.tm API via curl and executes rm -rf to clean up temporary test artifacts.- [EXTERNAL_DOWNLOADS]: Fetches Playwright testing tools and MCP servers from the npm registry at runtime. It also recommends the use of external plugins from well-known repositories associated with Vercel and Supabase.- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it ingests data from untrusted or external sources.
Ingestion points: The agent reads task requirements from docs/task/*.md and captures accessibility snapshots or screenshots of external web pages.
Boundary markers: No explicit delimiters or instructions are used to separate untrusted data from system instructions.
Capability inventory: The skill can execute shell commands (npx, npm, bash), control a web browser, and modify the local file system.
Sanitization: There is no evidence of input validation or sanitization for the content read from task documents or web pages before it is processed by the model.

test