The Agent Skills Directory

[METADATA_POISONING]: The skill's metadata identifies the author as 'vercel', which contradicts the established author 'attentiondotnet'. This mismatch is deceptive and can lead to a misunderstanding of the skill's true origin.
Evidence: The YAML frontmatter in SKILL.md explicitly lists author: vercel.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to read and analyze user-provided code files, creating a surface where malicious instructions within those files could override the agent's instructions during the review process.
Ingestion points: The skill ingests untrusted data from files provided via the $ARGUMENTS parameter as described in references/guidelines.md.
Boundary markers: The instructions lack explicit boundary markers or directives for the agent to treat file content as data rather than instructions.
Capability inventory: The skill utilizes file reading and text output; it does not demonstrate command execution, network exfiltration, or persistence capabilities.
Sanitization: There is no evidence of sanitization or content validation for the files being analyzed.
[EXTERNAL_DOWNLOADS]: The skill optionally retrieves updated guidelines from the official Vercel Labs repository on GitHub.
Evidence: SKILL.md notes an optional fetch from https://raw.githubusercontent.com/vercel-labs/web-interface-guidelines/main/command.md.

web-design-guidelines