Loading ToolBench
Single tool 'search_creatives' has a complete input schema with 10 parameters, all typed and described. However, the tool description is generic ('Search AdMapix advertising creatives and return raw data') and lacks context on WHEN to use it or WHAT it returns. No output schema is documented. The archived location (docs/archive/deep-research-agent/agent/tools.py) and lack of integration into the actual MCP server codebase (admapix_mcp/__main__.py not shown) suggest this tool definition may not be actively registered. Parameter descriptions are present but terse (e.g., 'Creative type codes' without explaining the binary encoding '100'/'010'/'001'). No error handling guidance, no recovery hints, and no mention of rate limits or API prerequisites.
Search AdMapix advertising creatives and return raw data.
Tool definition location is archived and not visible in active MCP server code. Cannot confirm tool is registered in the actual fastmcp server implementation.
No output schema documented. LLMs cannot plan downstream calls or extract required fields (e.g., creative IDs for follow-up actions).
Tool description is generic and lacks WHEN/WHY guidance. 'Search AdMapix advertising creatives and return raw data' does not explain when to call this vs. other discovery tools or what 'raw data' contains.
Parameter 'creative_team' uses opaque codes ('100', '010', '001') without clear explanation. Description says 'Creative type codes' but does not explain the binary encoding or why these specific values.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2026-07-28+ | v2 |
No error handling guidance. Tool implementation catches exceptions and returns JSON errors, but descriptions do not tell LLMs what to do on failure (retry, ask user, abort).