patch-skill

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.80). The skill explicitly fetches and decodes files from arbitrary upstream GitHub repos (e.g., gh api "repos///contents/skills/" and reading the .content), and those third-party repo files are parsed and diffed to decide what to include in commits/PRs, so untrusted repository content can influence agent decisions and actions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill instructs running "npx cyber-skills@ skill source " at runtime, which fetches and executes remote npm package code (cyber-skills) and therefore is a runtime external dependency that executes remote code.