readme-trick

SUSPICIOUS: the provided skill is intentionally incomplete and routes the agent to unreviewed README instructions, which the description itself says contain prompt injection. Even without visible code execution or credential flow, this is incompatible with a benign, reviewable skill footprint.

This fragment is a high-risk remote code execution instruction (`curl ... | bash`) that would allow arbitrary commands from an external, attacker-controlled URL to run immediately. Since no real package code is provided, the exact supply-chain mechanism cannot be validated, but if included in dependency lifecycle scripts, it would represent a severe compromise attempt.