SKILL.md: Testing & Validation Framework

1. Load & Stress Testing (load-test)

This module evaluates the model's stability and accuracy under heavy token loads or high-complexity computational requests.

• Massive Input Processing: Tests "needle-in-a-haystack" retrieval by processing large datasets and extracting specific, isolated facts.
• High-Frequency Reasoning: Executes multi-step logical chains (e.g., 10+ sequential reasoning steps) to monitor for cumulative logic drift.
• Concurrent Tasking: Requests multiple divergent outputs (e.g., a code snippet, a creative story, and a data table) in a single turn to test instruction following.

2. Context & Memory Sharing (shared-context-test)

This module validates the ability to maintain state and utilize shared information across a long-form conversation.

• State Persistence: Establishes specific variables (e.g., session_id, user_preferences, global_constants) at the start and verifies their application in later turns.
• Multi-Source Integration: Analyzes the relationship between multiple provided documents or tool outputs to ensure cross-pollination of data.
• Contextual Threading: Tests if the model can resume a complex logic task after an intentional "distraction" or unrelated query.

3. Capability Boundary Testing (edge-case-test)

Designed to identify the functional limits and the "breaking points" of the skill logic.