The Agent Skills Directory

[PROMPT_INJECTION]: The skill accepts user input for ingredients and equipment which are used to select recipes. Although these inputs are interpolated into text outputs, the skill's logic is based on static keyword matching and does not provide a mechanism for these inputs to override system behavior.
Ingestion points: user_request, context.inventory, and context.equipment in src/index.js.
Boundary markers: Output is segmented into defined artifact blocks (brief, table, plan).
Capability inventory: No dangerous capabilities (file writes, network requests, or shell execution) are exposed to the logic processing user data.
Sanitization: Inputs are processed via normalizeLanguage and topKeywords utilities before being used in logic.
[COMMAND_EXECUTION]: The inclusion of a Python script scripts/validate_operations_artifacts.py involves the use of subprocess.run. This script is a development-time utility used to verify the schema of output files against an internal shared validation tool. As it targets a known local path within the vendor's directory structure, it is considered a safe operational tool.

kitchen-commander