rules-lawyer

Pass

Audited by Gen Agent Trust Hub on Oct 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches official Magic: The Gathering Comprehensive Rules (TXT) and MTGJSON bulk data from remote sources. These resources are well-known community and official endpoints for game metadata.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data (MTG rules and card rulings) which constitutes an ingestion surface for potentially untrusted content.
  • Ingestion points: Content is ingested into the agent context via rules-lookup and rulings-lookup tools in SKILL.md.
  • Boundary markers: The skill includes strong instructional constraints ("The Iron Rule") requiring the agent to cite specific rule numbers and provide verbatim quotes, which acts as a structural boundary.
  • Capability inventory: The agent has access to Bash for CLI operations, Agent for spawning sub-agents, and standard Read/Write tools.
  • Sanitization: No explicit sanitization of the downloaded rules text is performed before it is read into the context.
  • [DYNAMIC_EXECUTION]: Employs the Agent tool to delegate complex rules reasoning to specialized sub-agents. These sub-agents are pre-loaded with specific slices of the Comprehensive Rules text extracted via Read and grep operations.
  • [COMMAND_EXECUTION]: Utilizes CLI tools (rules-lookup, rulings-lookup, download-rules) to interact with the local environment and manage rules data. These tools are part of the internal mtg_utils package mentioned in the compatibility requirements.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 15, 2026, 05:53 PM