Lean and assumes Claude's competence — no padding about what a tool/API is; every section (Top 7 Mistakes, Three-Step Registration, Critical Requirements) earns its place with high information density. Not level 2 because there is no unnecessary explanation to tighten.

Provides fully executable, copy-paste-ready artifacts: a complete Python tool class, a complete JSON configuration, and runnable bash verification/test commands. Not level 2 because nothing is pseudocode or missing key details.

The Three-Step Registration sequence has explicit validation checkpoints — a verification script asserting each of the 3 steps, plus the MANDATORY `test_new_tools.py ... -> 0 failures` feedback loop — matching the score-3 anchor with feedback loops for error recovery.

Body is a clear overview pointing one level deep to real, existing bundle files (testing-guide.md, advanced-patterns.md, implementation-guide.md, tool-improvement-checklist.md), each clearly signaled in a References section; content is appropriately split rather than inlined.

Lists multiple concrete actions — "creating tool classes, JSON configurations, validation, error handling, and test examples" — matching the score-3 anchor for multiple specific concrete actions; it goes beyond naming a domain to enumerating deliverables.

Explicitly answers both what ("Create new scientific tools ... with proper structure, validation, and testing") and when via a clear "Use when users need to add tools to ToolUniverse ..." clause, matching the score-3 anchor.

Covers natural phrasing a user would actually say — "add tools to ToolUniverse", "implement new API integrations", "create tool wrappers", "contribution guidelines" — giving good coverage of likely trigger variations.

Scoped to the ToolUniverse framework with distinct, domain-specific triggers ("tool wrappers for scientific databases/services", "ToolUniverse contribution guidelines"), making it unlikely to fire for unrelated skills.