The Agent Skills Directory

[SAFE]: The skill's operations are strictly limited to Magic: The Gathering deck building and optimization. It follows a structured, multi-phase process with built-in verification steps to ensure accuracy and safety.
[EXTERNAL_DOWNLOADS]: The skill fetches card metadata, oracle text, and pricing from well-known and trusted community services, including Scryfall and EDHREC. It also includes a utility for retrieving strategy research from the web.
[DATA_EXFILTRATION]: While the mtga-import utility accesses the local MTG Arena Player.log file, it does so solely to extract game-specific collection and wildcard data for the user's benefit. No data is sent to unauthorized external destinations.
[COMMAND_EXECUTION]: The skill executes local Python scripts via uv run. These scripts are part of the skill's internal logic for parsing deck lists, auditing legality, and calculating mana base health.
[INDIRECT_PROMPT_INJECTION]: The skill represents an indirect injection surface as it processes external deck lists and web-based strategy articles.
Ingestion points: External URLs via web-fetch and user-provided files via parse-deck.
Boundary markers: The instructions mandate the use of the Write tool for structured JSON intermediates, avoiding raw shell interpolation.
Capability inventory: Execution of local Python scripts, file system writes within the workspace, and network access to MTG-related APIs.
Sanitization: Input is passed through specialized parsing scripts that normalize data into canonical JSON schemas before LLM processing, significantly mitigating the risk of embedded instructions reaching the model's control flow.

deck-wizard