The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/_skill_cli.py includes a run_command function that uses subprocess.run to execute system commands. This is a standard utility for the skill's operations but establishes a capability for command execution on the host system.
[EXTERNAL_DOWNLOADS]: The SKILL.md file specifies the use of @playwright/mcp@latest via npx, which involves downloading and executing the latest version of the package from the npm registry at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill performs visual verification by navigating to and screenshotting a local Vite development server, creating an attack surface for indirect prompt injection.
Ingestion points: Data is ingested from http://localhost:5173 via browser navigation and screenshots as described in SKILL.md.
Boundary markers: The skill does not provide explicit delimiters or instructions to the agent to ignore or isolate potentially adversarial content within the rendered web pages.
Capability inventory: The agent is empowered to modify source code (Vue/CSS) and execute commands via utility scripts based on its interpretation of visual data.
Sanitization: The skill does not implement sanitization or validation of the rendered content before the agent processes it for decision-making.

visual-verification