Loading ToolBench
MCP server for lightweight work-first task handoffs across coding sessions.
Strong foundation with 17 well-named tools following verb_noun convention (get_, list_, start_, resume_, set_, read_, append_, update_, amend_, create_). All tools have descriptions (avg ~120 chars, within 10-1024 baseline). Input schemas present for all tools with proper type definitions and enums. However, output schemas are partially documented, some tools return structured content but lack explicit schema declarations in the registration code. Error handling is minimal; no recovery guidance or actionable error messages visible. Security considerations present (no secrets in params) but no permission gates or audit trail patterns. Tool composition is sound with clear single responsibilities.
Amend metadata fields of an existing log entry.
Append a session log entry for the current work or project.
Append a lightweight note to the notes.md artifact for a work item.
Create or update a design.md artifact for a work item.
Create or update a plan.md artifact for a work item.
Create or update a spec.md artifact for a work item.
Create or update a summary.md artifact for a closed work item.
Output schemas not explicitly declared in tool registration. Tools return structuredContent but lack formal outputSchema definitions in registerTool() calls, making it unclear to clients what fields to expect.
No error handling or recovery guidance. Tools lack actionable error messages. When a work_id is invalid or a file operation fails, there is no guidance for the LLM on what to do next (retry, lookup, ask user).
Document artifact tools (create_design_doc, create_plan_doc, create_spec_doc, create_summary_doc) have minimal descriptions (~75 chars). Descriptions should clarify when to use each artifact type and what content is expected.
Inferred effective spec: 2025-06-18+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 61 | 2025-06-18+ | v2 |
Return the inferred project roots, canonical log paths, workdocs root, and current active work state for this session.
Get open log entries (deprecated; prefer list_works for work discovery).
Get recent session log entries, optionally filtered by work or project-wide.
List work items for the current project. Prefer this over open-thread log discovery when the user asks what is still open.
Read the full context for a work item, including artifacts, recent logs, and re-entry briefs.
Set the active work for this session by work id or query. Use this before reading work context or writing work-scoped logs.
Update the impact level of a work item.
Update the lifecycle status of a work item.
Create a new work, generate a short work id, persist work metadata, and mark it as the active work for this session.
Update the lifecycle status of a log entry.
No permission gates or audit trail patterns. Tools that modify state (start_work, append_session_log, create_design_doc) lack permission checks or logging of who called what and when.
get_open_threads marked deprecated but still exposed. Description says 'prefer list_works' but tool remains callable. Consider removing or adding a deprecation notice with migration path.