restore-credentials
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [METADATA_POISONING]: The skill's YAML frontmatter includes a future 'last-updated' date (2026-08-21) and claims 'Google LLC' as the author, which differs from the actual author context 'talissonvitorino.' This is likely a result of mirroring official documentation.
- [INDIRECT_PROMPT_INJECTION]: The skill involves the agent reading and modifying sensitive project files like the Android manifest and build configuration.
- Ingestion points: Reads
AndroidManifest.xmlandbuild.gradlefrom the user's project environment. - Boundary markers: The instructions provide clear architectural tiers and implementation guidelines to constrain agent behavior.
- Capability inventory: The skill workflow includes writing Kotlin classes and updating project-level XML and Gradle configuration files.
- Sanitization: The skill utilizes static implementation templates.
Audit Metadata