sync-provider

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

• Third-party content exposure detected (high risk: 0.95). The workflow ingests outsider-authored free text from upstream provider repositories via the runtime-generated diff artifacts (e.g., find .diffs/<commit_hash>/ -exec cat {} and ls -R .diffs/<commit_hash>/), where .diffs/<commit_hash>/*.txt contains upstream changes authored by third parties.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill fetches upstream repository content at GitHub URLs (e.g., https://github.com/ben-vargas/ai-sdk-provider-claude-code and the generic https://github.com/owner/repo-name) during runtime (via gh / scripts/git-diff.ts) and mandates submitting those fetched diffs/files to the external Pal MCP refactor tool (model "gemini-3.0-pro"), which injects remote content into model prompts and thus directly controls agent instructions.