Loading ToolBench
MCP server for discovering and converting Claude skills
This server demonstrates solid foundational quality with 8 well-named tools covering a coherent skill discovery and conversion workflow. All tools have descriptions and explicit input schemas with typed properties. However, there are meaningful gaps in parameter descriptions, output schema documentation, and error handling guidance that prevent a higher score. The tool set is cleanly composed with no redundancy, and naming follows verb_noun convention consistently. Parameter constraints (enums, required fields) are present where relevant. The primary limitation is incomplete documentation of what each tool returns and minimal error recovery guidance.
Convert skill to Kiro power format
Convert skill to Kiro steering file format
Fetch raw skill content from GitHub
Get trending or top-installed skills
Complete import workflow (fetch + validate + convert)
List all available skills from skills.sh with pagination
Search for skills by keyword
Output schemas not documented. No tool specifies what fields are returned or their types. LLMs cannot plan downstream calls or extract relevant data without knowing the response structure.
Parameter descriptions are minimal. Examples: 'page' → 'Page number (default: 1)' lacks context on usage; 'identifier' → 'Skill name or owner/repo format (required)' is underspecified (what is the exact format? examples?). LLMs need explicit guidance on valid input formats.
No error handling guidance. Tools return no indication of what to do if a skill is not found, validation fails, or conversion fails. Error responses should guide recovery (e.g. 'Skill not found. Try search_skills() first').
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 67 | 2026-07-28+ | v2 |
Validate skill content for security issues
Pagination and result limits undocumented. list_skills accepts page/pageSize but no indication of total count, next_cursor, or whether results are sorted. No clear behavior when querying beyond available pages.
Tool descriptions lack context on when to use each tool vs. alternatives. E.g., search_skills vs. get_leaderboard, when should an LLM pick one over the other? What is the distinction in use cases?
validate_skill tool provides no output spec. Does it return a boolean, a detailed report, or a list of issues found? How should the LLM interpret success vs. failure?
Conversion tools (convert_to_steering, convert_to_power) have vague descriptions. What is 'Kiro' format? Why two formats? When should an LLM use one vs. the other? Descriptions should answer these questions.