analyze-deps-guide

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

• Third-party content exposure detected (high risk: 0.85). In Phase 3 the skill performs runtime web fetching of outsider-authored free text from GitHub Releases/CHANGELOG.md and WebSearch migration guides (e.g., https://api.github.com/.../releases and https://raw.githubusercontent.com/.../CHANGELOG.md), which is then parsed and included in the generated report/LLM context.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly fetches and parses external release notes/changelogs at runtime—e.g. https://api.github.com/repos/{owner}/{repo}/releases and https://raw.githubusercontent.com/{owner}/{repo}/main/CHANGELOG.md (with master fallback)—and injects that content into its analysis/report, meaning remote content fetched during runtime can directly influence the agent's outputs.