Loading ToolBench
Fast, deterministic, zero-API linter for AI agent prompt assets — SKILL.md, sub-agents, MCP tool descriptions, slash commands, and system prompts
Single tool 'lint_prompt' has a clear action verb and reasonable description (67 chars). Input schema is present with three parameters, all typed and described. However, the tool lacks output schema documentation, critical for LLMs to understand what fields to expect from linting results. Parameter descriptions are adequate but could be more prescriptive about constraints and error cases. No error handling guidance provided. The tool is read-only (low risk) but lacks the structured output documentation expected of production tools.
Lint a prompt asset and return findings.
Output schema not documented. LLMs cannot infer what fields the linting response contains (e.g., findings array structure, severity levels, line/column format). This forces agents to guess at response shape and risks parsing failures.
Parameter 'kind' accepts free-form string with enum values listed only in description ('skill', 'subagent', 'command', 'mcp-tool', 'prompt'). Should be declared as JSON Schema enum constraint, not prose. LLMs cannot parse prose enums reliably.
Parameter 'fmt' similarly lacks enum constraint. Prose description lists 'json', 'text', 'github', 'sarif' but schema should enforce these as enum values. Prevents LLM hallucination of invalid formats.
No error handling guidance. Tool description does not explain what happens on invalid input (malformed YAML, unknown document kind, etc.) or how the agent should recover. Missing recovery patterns.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 55 | 2026-07-28+ | v2 |
Tool description does not explain when to call lint_prompt vs. other linting tools, or what 'findings' structure contains. Lacks context for tool selection and downstream processing.