The Agent Skills Directory

[SAFE]: The skill instructions define a read-only subagent focused on quality assurance and security auditing of Android features. The 'adversarial auditor' and 'red-team' language used throughout the instructions is context-appropriate for its intended function as a validator and does not constitute a prompt injection attack against the agent's safety guidelines.
[SAFE]: The skill includes comprehensive security checklists for auditing Android application code, covering intent redirection, secret management, and input validation, which are best practices for the target domain.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input in the form of code diffs and implementation files produced by other agents. It possesses capabilities to execute local commands like ./gradlew and android-cli. While this presents an attack surface for indirect prompt injection, it is the primary purpose of the skill to audit such content, and the risk is considered low within this controlled development context.

workflow-validator