Loading ToolBench
Comic-style ASCII art with speech bubbles, bold banners, and action effects. Uses categorical composition of aesthetic functors with zero rendering dependencies.
ASCII Comic Generator has 8 tools with complete input schemas and descriptions, but quality is uneven. All tools have descriptions (10-100 chars, mostly adequate), and schemas are present with proper types and enums. However, descriptions are often terse (e.g., 'Create a speech bubble with text and optional speaker name' is 56 chars, functional but minimal). Parameter descriptions are present but generic ('The text content of the speech bubble'). No output schemas are documented, forcing LLMs to guess what fields are returned. Error handling is absent, no guidance on what happens if a template doesn't exist or rendering fails. The tools are well-named (verb_noun pattern) and properly constrained with enums, but lack the depth needed for confident LLM selection in complex scenarios. Composition is sound (each tool does one thing), but missing idempotency guarantees and recovery guidance.
Apply ASCII shading to text or ASCII art
Compose multiple ASCII art elements into a comic panel
Create an action effect text with visual emphasis
Create a bold banner with text
Create large ASCII text using figlet fonts
Create a speech bubble with text and optional speaker name
List all available ASCII art templates
Render an ASCII art template for a subject
No output schemas documented. LLMs cannot predict what fields are returned (e.g., does render_template return just ASCII text, or also metadata like dimensions?). Forces guessing and downstream parsing errors.
No error handling guidance. If render_template('invalid_subject') fails, the LLM has no recovery path. Should return 'Subject not found. Available: dog, cat, bird, ...' to enable self-correction.
Parameter descriptions are generic and lack constraints. 'The text content of the speech bubble' does not specify max length, character restrictions, or newline handling. 'Width of the speech bubble in characters' lacks min/max bounds.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2026-07-28+ | v2 |
compose_comic_panel 'elements' parameter is an array with no item schema. LLMs cannot determine what structure each element should have (string? object with type and content?).
No idempotency guarantees. If an LLM retries create_speech_bubble with identical params, will it produce the same output or a different rendering? Agents need to know if retries are safe.