Loading ToolBench
The server defines 3 tools with reasonable naming (verb_noun pattern) and basic descriptions. However, parameter descriptions are minimal or missing entirely, input schemas lack full type specifications for all parameters, and output schemas are undocumented. Error handling exists but provides minimal guidance. The server reads from static files and has no destructive operations, which mitigates some risk, but the interface lacks the structure and clarity required for high-confidence agent use.
Return a full, Markdown formatted CV/resume for Tim Cave, including work experience, education, skills, and any other relevant information.
Get detailed information about a specific project. Returns full contents of the associated info file if it exists. Use get_project_list with id_only=True to get a list of project IDs to use as input for this function.
Get a list of projects working on by Tim Cave, including descriptions, status and key skills.
Parameter descriptions are missing or minimal. 'id_only' in get_project_list and 'outcomes_only' in get_project_info lack description text in the visible function signatures. LLMs cannot infer when to set these flags without explicit descriptions.
Output schemas are completely undocumented. All three tools return strings, but LLMs need to know the structure of those strings (is it markdown? JSON? plain text? field names?). Without documented output schemas, agents cannot reliably parse results or chain tools.
Parameter 'project_id' in get_project_info lacks format/constraint documentation. The description says 'The ID of the project' but does not specify what constitutes a valid ID, expected format, or how to discover valid IDs (though it does mention calling get_project_list first, good).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 69 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 46 | - | v1 |
Error messages lack recovery guidance. When CV file is missing, the response is 'CV file not found, bad luck.' This is humorous but unhelpful to an LLM, it should suggest next steps (e.g., 'CV file not found. The server may be misconfigured.'). Similarly, 'Project not found' should suggest calling get_project_list to verify the ID.
Tool descriptions do not specify output content or format. 'Return a full, Markdown formatted CV' is good, but does not describe what fields or sections are included, expected length, or how to use the output downstream. Agents need to know if the CV is suitable for emailing, summarizing, or extracting specific skills.