Loading ToolBench
A custom MCP server for managing, building, editing, and customizing projects with AI. Maintains a persistent project context document that captures tech stack, architecture, conventions, key files, active tasks, goals, and architectural decisions. Provides prebuilt prompt templates for common operations.
Server has 20 tools with consistent naming (verb_noun pattern) and descriptions present for all. However, input schemas lack critical details: most parameters have only type and description, missing constraints (enums, min/max, patterns). Output schemas are entirely undocumented, LLMs cannot predict return structures. Error handling is absent from visible code. Parameter descriptions are adequate (avg ~80 chars) but lack actionable guidance on format/range. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels. Composition is sound (single responsibility per tool), but chaining support is weak, tools don't consistently return IDs needed by downstream operations.
Define a milestone with a target date and associated features.
Scan files and construct a complete mapping of dependencies and references in the project.
Compare current project context and tasks against PRD requirements to check alignment.
Mark the current phase as completed and optionally advance to the next phase.
Generate a session summary capturing what was done, what's next, and current project state for handoff to another AI.
Generate a complete aggregated Project Brain dashboard showing current status, phase, active/blocked tasks, recent changes, and recommendations.
Output schemas completely undocumented. No tool returns a documented structure, LLMs cannot predict fields, types, or pagination. Violates pattern:tool and pattern:response-shaper.
Input parameters lack constraints. No enums for 'document' param in get_prd/update_prd (should be enum: PRD|VISION|SCOPE|REQUIREMENTS). No min/max for numeric fields. No regex patterns. Violates pattern:constrained-input.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 53 | 2026-07-28+ | v2 |
Get the current active phase and its status.
Retrieve the most recent AI handoff document so a new AI session has full context.
Retrieve a PRD document (PRD, VISION, SCOPE, REQUIREMENTS). Returns the content of the specified document.
Calculate direct and transitive downstream files affected if a given file is modified.
List all AI session handoffs with dates.
Show all milestones and their status.
List all phases and their status (completed, active, pending).
Intelligently calculate and recommend the next best task to work on based on phase, priority, dependencies, and file hints.
Mark current phase as complete and advance to the next phase in sequence.
Set the current active development phase. Only tasks in this phase are considered 'in scope'.
Start a development session. Loads state, PRD, active tasks, and latest handoff to generate a detailed Session Brief for the AI.
Summarize key requirements from PRD documents into a concise bullet list.
Create or update a PRD document (PRD, VISION, SCOPE, REQUIREMENTS).
Scan the codebase and validate imports/technologies against the project's architectural guidelines and forbidden tools.
No tool annotations despite clear risk levels. WRITE tools (generate_handoff, set_phase, complete_phase, next_phase, add_milestone, update_prd) lack destructiveHint. READ_ONLY tools lack readOnlyHint. Violates pattern:tool-annotation.
No error handling guidance visible. Tools return no recovery hints, error classification, or actionable messages. Violates pattern:recovery-guide and pattern:error-classification.
Pagination not visible for list_* tools. list_handoffs, list_phases, list_milestones lack limit/offset/page parameters and total_count in response. Violates pattern:paginated-result.