The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill directs the agent to use npx --yes notugly, which automatically downloads and executes code from an unverified package on the NPM registry at runtime.
[COMMAND_EXECUTION]: Instructions include executing shell commands such as npx --yes notugly steal [URL] and npx notugly fix [hex] [hex]. These commands incorporate user-provided inputs (URLs and color codes) directly into a shell environment, which is susceptible to command injection attacks if the input is not strictly validated.
[EXTERNAL_DOWNLOADS]: The skill relies on fetching external Node.js packages via npx without version pinning or integrity checks, exposing the environment to supply chain risks or malicious package updates.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data with high-capability tools.
Ingestion points: User-provided website URLs and screenshots (processed via steal and spec commands).
Boundary markers: The skill lacks explicit instructions or markers to distinguish between legitimate stylesheet data and potential malicious instructions embedded in CSS or HTML metadata.
Capability inventory: The agent is authorized to execute shell commands (npx) and read local files (artifacts/decks).
Sanitization: There is no evidence of sanitization or escaping of external inputs before they are passed to shell commands.

brand-guidelines