edge-to-edge

Warn

Audited by Gen Agent Trust Hub on Oct 5, 2026

Risk Level: MEDIUMMETADATA_POISONINGCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [METADATA_POISONING]: The skill's metadata incorrectly identifies 'Google LLC' as the author and provides a future last-updated date ('2026-08-24'). These discrepancies from the actual author identity and current time are deceptive and can lead users to grant undue trust to the instructions based on a false sense of institutional authority.\n- [COMMAND_EXECUTION]: The skill instructs the agent to execute the ./gradlew build command to verify the project status. While this is a standard developer workflow, it requires the agent to have shell execution capabilities, which could be misused if project context is compromised.\n- [INDIRECT_PROMPT_INJECTION]: The skill directs the agent to analyze arbitrary local source code (Activities and AndroidManifest.xml) and subsequently execute build commands, creating a surface for instructions hidden in project files to influence agent behavior.\n
  • Ingestion points: Project source files, specifically all Activity classes and the AndroidManifest.xml file, as defined in Step 1 and Step 2 of SKILL.md.\n
  • Boundary markers: None identified; the instructions do not include specific delimiters or warnings to ignore instructions embedded within the analyzed code.\n
  • Capability inventory: Execution of the ./gradlew build shell command (SKILL.md).\n
  • Sanitization: None identified; there are no instructions for the agent to validate or sanitize the content of the source files before processing them.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 5, 2026, 09:27 PM