Provider-neutral, local-first comic production pipeline with MCP integration for agent-based comic generation, lettering, typesetting, and PDF export
Comic Sol has 16 tools with complete input schemas and descriptions, but quality is inconsistent. Most tool descriptions are adequate (50-150 chars), but several lack actionable context for LLM selection. Parameter descriptions are present but often generic ('Project identifier' repeated 16 times). No output schemas documented. Error handling guidance absent. Tool naming follows verb_noun convention (init, validate, resume, promote_attempt) but some names are vague (doctor, transition). No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications in metadata. Composition is reasonable, tools chain logically (init → validate → transition → letter_project → compose_pages → export_pdf), but no pagination, batch operations, or recovery guidance.
Compose lettered panels into final pages
Validate project structure and report health status
Export composed pages to PDF format
Finalize a project and mark it as complete
Query the current image generation capability status and features
Initialize a new comic project with source material and configuration
Apply lettering and typesetting to project panels
Promote a generation attempt to the active panel image
No output schemas documented. LLMs cannot infer what fields to expect from tool responses, forcing them to guess at downstream field names and risk broken tool chains.
Parameter descriptions are generic and repetitive ('Project identifier to validate', 'Project identifier', 'Project identifier'). LLMs need context on what each tool does differently and when to select it over similar tools.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 52 | 2026-07-28+ | v2 |
Read the complete project manifest with all configuration and state
Read the current status and metadata of a project
Record a generation attempt for a panel with specified kind and metadata
Record a manual override decision for a panel with reason
Render a human-readable report of project status and issues
Resume a paused or failed project from a specific stage
Move a project to a new workflow status with optional warning
Validate project at a specific workflow stage
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk classifications in metadata. LLMs cannot determine which tools are safe to retry or have side effects.
No error handling guidance. Tools lack recovery hints (e.g., 'If project not found, call list_projects() first'). Agents cannot self-correct on failures.
get_image_capability has confusing schema: 'status' enum with values 'available'/'unavailable', but also 'name' with const='agent-image-generation'. This looks like a capability query response schema misplaced as input parameters. Unclear what LLM should pass.