Browse, search, and pull AI agent skills from any MCP-compatible client
OpenBooklet MCP server has 8 tools with consistent naming (verb_noun pattern: search_, get_, pull_, trending_, resolve_) and complete input schemas using Zod. All tools have descriptions (avg 120 chars, within 10-1024 baseline). However, output schemas are NOT documented in the source code, only TypeScript interfaces exist internally but are not exposed to the MCP client. Parameter descriptions are present but generic (e.g., 'Skill name/slug' lacks format guidance). No error handling guidance, no recovery hints, and no tool annotations (readOnlyHint missing despite all tools being read-only). Tool composition is sound (each does one thing), but output documentation is a critical gap for LLM reasoning.
Get full details about a specific skill including metadata, stats, and publisher info.
Pull an MCP server from OpenBooklet. MCP servers are either personas (load into system prompt) or knowledge packs (retrieved into context). The response includes the subtype.
Pull the raw content/instructions of a skill. This is the actual content that gets injected into an agent's context.
Pull the raw instructions of a workflow from OpenBooklet.
Pull a workflow and ALL its locked dependency skills and MCP servers as a single bundle. This is the recommended way for agents to load a complete workflow — one call gets everything.
Resolve the full dependency tree for one or more skills. Returns a topologically-sorted install order including all transitive dependencies.
Output schemas not documented. TypeScript interfaces (SkillDetail, WorkflowDetail, McpServerDetail, TrendingResult) exist in code but are not exposed to MCP clients. LLMs cannot plan downstream tool calls or extract fields without knowing response structure.
No tool annotations. All 8 tools are read-only (no side effects), but readOnlyHint is not set. LLMs cannot distinguish safe tools from destructive ones, risking unnecessary caution or unsafe assumptions.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 69 | 2026-07-28+ | v2 |
Search the OpenBooklet registry for skills, workflows, or MCP servers. Returns matching assets with name, description, badge, and pull count.
Get the top trending assets this week on OpenBooklet. Supports filtering by type.
No error handling guidance. Tools call external API (apiFetch, apiText) but error responses are generic ('API error 404: ...'). No recovery hints, no categorization (retryable vs user-fixable), no suggestions for alternative tools.
Parameter descriptions lack format/constraint details. 'Skill name/slug (e.g. code-review-pro)' includes example but no regex, length limits, or character restrictions. LLMs may pass invalid formats.
No pagination guidance in search_skills. Limit defaults to 10 and caps at 50, but no mention of total count, next_cursor, or how to iterate large result sets. LLMs may assume all results fit in one call.