Experimental agentic novel generation CLI with world-building, character planning, scene writing, and skill management capabilities
Novel-Claude exhibits severe definition quality gaps across all dimensions. Tool names lack consistent verb-noun structure (e.g., 'expand', 'world', 'blueprint' are vague nouns, not action verbs). Descriptions are present but generic and lack LLM-optimization guidance. Most critically, NO input schemas are visible in the provided source code, only Click argument/option decorators are shown, not JSON Schema definitions required for MCP tool registration. Parameter descriptions exist in Click help text but are not formalized in structured schemas. Output schemas are completely undocumented. Error handling is minimal (basic ClickException usage). The codebase appears to be a CLI tool being retrofitted as MCP, not a native MCP implementation.
Audit one stage or one chapter for consistency
Build a Zhipu Batch API JSONL request file
Submit a JSONL file to the Zhipu Batch API
Poll and merge a Zhipu Batch API job
Regenerate character, scene, and organization cards
Show file contents
Change current directory
NO JSON Schema definitions visible for any tool. All 30 tools show Click decorators but no MCP-compliant input schemas.
Vague tool names lacking action verbs. 'expand', 'world', 'blueprint' are nouns, not verbs. Should be 'regenerate_outline', 'regenerate_world', 'regenerate_blueprint' or similar. LLMs rely on verb-noun naming to infer intent.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 33 | 2026-07-28+ | v2 |
Regenerate the story outline from initialized setting cards
Find files matching pattern
Run the complete world-building pipeline from a LOGLINE
List directory contents
Plan all volumes, or one volume via `plan 1` / `plan --volume 1`
Create a new project
Delete a project
Show current project info
List all projects
Switch to a different project
Print working directory
Reindex completed chapters into the enabled RAG Skill
Use the configured model to review and rewrite selected files
Set a configuration value
Show current configuration
Generate a new skill from a request
Disable a skill
Enable a skill
List, enable, disable, reload, or generate Skills
Reload one or all skills
Track entity state changes for a generated chapter
Regenerate the world setting from the story outline
Generate chapters with progressive saving and enabled Skills
No output schemas documented. Tools return dict with 'error', 'message', 'output' keys, but no formal schema definition for downstream tool chaining. Agents cannot plan multi-step sequences without knowing return structure.
Minimal error handling. Tools raise ClickException with generic messages. No recovery guidance, no error classification (retryable vs fatal), no actionable next steps for LLM.
Tool definitions appear to be inferred from Click CLI code, not explicitly registered as MCP tools. No visible MCP server initialization, tool registration, or protocol handshake. This is a CLI tool, not an MCP server.