Loading ToolBench
MCP server providing task and project management tools for AI agents
Server provides 5 tools with complete input schemas and descriptions. Naming follows verb_noun convention (todo_manager, priority_sorter, time_estimator, standup_generator, burndown_calculator). Descriptions are present and substantive (100-200 chars typical). However, critical gaps exist: (1) No output schemas documented, LLMs cannot predict return structure for downstream chaining; (2) Parameter descriptions lack actionable constraints (e.g., 'reach' in RICE has no range guidance, 'buffer_percent' lacks min/max); (3) No error handling guidance, tools fail silently with no recovery hints; (4) todo_manager combines 5 actions (add/complete/update/delete/list) in one tool, violating single-responsibility; (5) priority_sorter's task input schema is underspecified, which fields are required per method? Eisenhower needs urgent/important, RICE needs reach/impact/confidence/effort, but schema doesn't enforce method-specific requirements. Baseline: 94% of A+ tools have param descriptions; this server has descriptions but they lack depth.
Calculate sprint burndown chart data. Given tasks with story points and sprint duration, computes ideal vs actual burndown, velocity metrics, and completion forecast.
Sort and prioritize tasks using proven frameworks: Eisenhower matrix (urgent/important quadrants), RICE scoring (reach, impact, confidence, effort), or MoSCoW method (must/should/could/wont).
Generate a daily standup update from the current task list. Shows yesterday (recently completed), today (in progress), and blockers. Uses data from todo_manager.
Estimate task duration using PERT methodology. Break a task into subtasks, provide optimistic/realistic/pessimistic estimates for each. Returns weighted expected time plus configurable buffer.
Manage an in-memory task list. Add, complete, update status, delete, and list tasks. Supports filtering by status, priority, and tags. Persistent within session.
No output schemas documented. LLMs cannot predict return structure, breaking tool chaining and forcing agents to guess field names for downstream calls.
todo_manager violates single-responsibility: combines add, complete, update_status, delete, list into one tool. Should split into separate tools (create_task, complete_task, update_task, delete_task, list_tasks) for composability.
priority_sorter schema does not enforce method-specific required fields. Eisenhower requires urgent/important booleans, RICE requires reach/impact/confidence/effort numbers, Moscow requires moscow enum, but schema marks only id/title as required. LLM may pass incomplete task objects.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 51 | 2026-07-28+ | v2 |
Parameter descriptions lack actionable constraints. 'reach' (RICE) has no range guidance; 'buffer_percent' lacks min/max; 'impact' lacks 0.25-3 range in description (only in schema). Baseline: 100% of A+ tools specify ranges and formats in descriptions, not just schema.
No error handling or recovery guidance. Tools provide no hints for LLM on what to do if a call fails (e.g., 'Task not found, try list_tasks first'). Agents cannot self-correct.