mlb-closer-tracker

Pass

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [SAFE]: The skill's operations are entirely focused on gathering and analyzing MLB bullpen data. Its stated purpose aligns with its behavior, and it does not perform any sensitive system operations or unauthorized network communication.
  • [INDIRECT_PROMPT_INJECTION]: The skill has a data ingestion surface because it retrieves information from external websites via web search and direct URLs.
  • Ingestion points: The skill pulls data from RotoBaller, Closer Monkey, Baseball Savant, and team beat writers to determine relief pitcher roles.
  • Boundary markers: No explicit boundary markers or instructions to ignore instructions embedded in external content are present in the skill instructions.
  • Capability inventory: The skill is limited to writing markdown files to a local signals/ directory. It lacks capabilities for arbitrary shell command execution, sensitive file access (e.g., SSH keys, credentials), or network exfiltration of local data.
  • Sanitization: The skill processes external text for role scoring but does not perform explicit data sanitization. However, since the output is restricted to markdown reports, the risk of cross-context instruction execution is minimal.
  • Note: As data ingestion is the primary intended function for this research tool, this low-risk surface does not escalate the overall verdict.
  • [EXTERNAL_DOWNLOADS]: The skill references and fetches data from several external services. Per security guidelines, these sources (RotoBaller, Closer Monkey, Athlon Sports, Yahoo Sports, and MLB.com) are recognized as well-known technology or media services. No executable code or scripts are downloaded from these domains.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 21, 2026, 05:59 AM