pvp-winrate-analyzer

Pass

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • [SAFE]: The skill follows security best practices by providing guidelines for quoting SQL identifiers to prevent injection and requiring explicit user confirmation before performing write operations or modifying the local environment.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface as it processes user-provided event and property names to construct analysis commands.
  • Ingestion points: Event names, property names, and project IDs supplied in the 'Required Inputs' section.
  • Boundary markers: The instructions guide the agent to use structured clarifications if names are unresolved.
  • Capability inventory: Shell command execution via 'ae-cli' for data retrieval and export.
  • Sanitization: Explicit instructions for quoting SQL identifiers containing special characters.
  • [COMMAND_EXECUTION]: The skill utilizes a set of 'ae-cli' commands to interact with project data and metadata. These commands are integral to the skill's primary function of gaming data analysis.
  • [DYNAMIC_EXECUTION]: The skill includes instructions for handling missing Python dependencies through isolated virtual environments, which is a standard development practice. The 'Failure Handling' section specifies asking the user before creating environments or downloading packages.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 8, 2026, 03:22 AM
Security Audit — agent-trust-hub — pvp-winrate-analyzer