update-skills-from-changelog

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

• Third-party content exposure detected (medium risk: 0.30). SKILL.md describes fetching and reading a merged changelog file from the external elevenlabs-dx GitHub repository and then ingesting its contents to drive the update workflow, so outsider-submitted changelog text could be consumed at runtime.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill runtime explicitly fetches remote changelog and docs (e.g., using gh api "repos/elevenlabs/elevenlabs-dx/contents/fern/docs/pages/changelog/${CHANGELOG_DATE}.md?ref=main" and other elevenlabs.io/github URLs) and uses that fetched content to drive edits/PRs, so external content directly controls the agent's instructions.