Ancient wisdom for modern systems. MCP server providing Taoist principles for software development.
The server has 4 well-named tools with clear action verbs (get_*, search_*), and parameter schemas are present with type constraints and enums. However, descriptions are generic and lack LLM-optimized guidance on when/why to use each tool. Parameter descriptions are minimal. Output schemas are not documented in the code, clients cannot predict what fields they'll receive. Tool compositions are weak: there's no clear chain between search_wisdom and get_chapter, and philosophical context is isolated. Error handling is not visible in the provided code. The schema quality is moderate but output documentation is entirely absent.
Retrieve a specific chapter from The Way of Code (1-81)
Retrieve philosophical context and background on The Way of Code
Retrieve principles organized by specific topics
Search through The Way of Code chapters by query string, with optional context and result limits
Output schemas not documented. Code snippet cuts off mid-implementation; cannot verify what fields tools return. LLMs cannot plan downstream tool calls or extract data without knowing response structure.
Parameter descriptions are trivial (under 20 chars). E.g., 'Chapter number (1-81)' and 'Search query string' repeat the schema constraint without explaining WHEN/WHY to use the param or what semantic the LLM should reason about.
Tool descriptions are generic (45-75 chars) and lack domain context. 'Retrieve a specific chapter' does not explain that chapters are numbered 1-81 by Taoist principle, when an LLM should prefer get_principles_by_topic (by semantic topic) vs get_chapter (by numeric reference), or how output relates to the broader wisdom resource.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 52 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
No tool composition strategy. search_wisdom can find relevant passages, but no guidance on how output IDs/chapter refs connect to get_chapter. An LLM cannot reason 'search first, then fetch full text' without explicit chaining hints in descriptions.
No error handling visible in provided code. No guidance on what happens if chapter number is out of bounds, search query is too broad, or topic enum is misspelled. LLMs need actionable error messages ('Chapter not found; valid range is 1-81') to self-correct.