mlb-opponent-profiler

Pass

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from Yahoo Fantasy Baseball pages, including team names, manager aliases, and transaction descriptions, which are then stored in markdown files within the context/opponents/ directory.
  • Ingestion points: Data is scraped from Yahoo league indices, team rosters, and transaction logs using the claude-in-chrome MCP tools.
  • Boundary markers: The templates provided in resources/template.md do not utilize explicit delimiters or 'ignore embedded instructions' markers for user-controlled strings like team_name or manager_alias.
  • Capability inventory: The skill possesses the ability to scrape arbitrary web content and write that content to local markdown files that are later consumed by five other agents (mlb-lineup-optimizer, mlb-category-state-analyzer, etc.).
  • Sanitization: There is no evidence of sanitization or filtering of the scraped text before it is written to the persistent context files, creating a potential vector for malicious managers in the fantasy league to influence downstream agent behavior via their team names or profiles.
  • [EXTERNAL_DOWNLOADS]: The skill navigates to and scrapes data from Yahoo domains (baseball.fantasysports.yahoo.com).
  • Evidence: The resources/methodology.md file specifies a scrape flow involving four distinct Yahoo Fantasy Baseball URLs.
  • Context: Yahoo is a well-known service, and this scraping behavior is the primary and intended purpose of the skill. This activity is considered safe under standard scope rules.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 21, 2026, 05:59 AM