The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for using the sops CLI tool to decrypt configuration files (components.sops.json) in specific development environments. This is a standard workflow for secret management and includes explicit guidance to avoid printing decrypted content to logs or the console.
[METADATA_POISONING]: The skill metadata identifies 'vercel' as the author, despite the skill being provided by 'darkmatter'. This reflects the content's origin (Vercel's public engineering guidelines) rather than a malicious attempt to deceive, as the author context acknowledges the 'darkmatter' association.
[INDIRECT_PROMPT_INJECTION]: As the skill is designed to assist in code refactoring and generation, it is theoretically susceptible to instructions embedded in the code it processes. However, no exploitable capability tiers (such as arbitrary network exfiltration or shell access) are exposed by this skill that would escalate this beyond a low-risk architectural factor.

vercel-react-best-practices