The Agent Skills Directory

[PROMPT_INJECTION]: Deceptive metadata in SKILL.md incorrectly identifies the author as "Google LLC," whereas the actual author is "talissonvitorino." This misrepresentation can lead to an inflated trust in the tool's authority and safety.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface. It ingests untrusted data from Android application source files and Google Play Store HTML, which is then interpolated into prompts for worker and critic sub-agents (e.g., in orchestrator.py and goal_data_safety.md).
Ingestion points: File content from the app_dir (targeted Android project) and HTML from the Play Store are read.
Boundary markers: The prompts generated for sub-agents lack explicit delimiters or instructions to ignore instructions within the analyzed code snippets.
Capability inventory: Sub-agents have file-writing capabilities, and the main orchestrator can execute Python scripts.
Sanitization: The scanner.py script attempts to mitigate this by stripping comments from XML and source files before processing.
[SAFE]: The skill fetches application metadata and data safety declarations from the official Google Play Store (play.google.com). This is a well-known and expected service for the skill's stated purpose of auditing Android apps.

play-policy-insights