search-context

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s runtime workflow uses gh api to discover and then clones and inspects arbitrary GitHub repositories, ingesting their README/description and code text via rg --json (with matched line text) into the generated manifest: scripts/search-context.mjs → discoverRepositories() → cloneRepository() → inspectRepository() → runLocalRepositoryProof() → renderManifest().

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

• Potentially malicious external URL detected (high risk: 0.80). This skill uses GitHub discovery and git clone at runtime to fetch repository content which is inspected and injected (short snippets and file references) into the generated manifest that will steer a coding agent — e.g., the example repo URL https://github.com/example/progress-ring-swiftui is referenced and represents such runtime-fetched content.