MCP server exposing the agent-tasks API as tools for MCP-capable clients
14 tools with complete input schemas and descriptions present. Naming follows verb_noun pattern (task_*, project_*, signals_*, workflow_*). Descriptions are adequate (50-150 chars typical) but lack LLM-optimized guidance on WHEN to use each tool vs alternatives. Parameters are typed with enums where appropriate (status, priority, outcome). Critical gap: output schemas are NOT documented, LLMs cannot plan downstream calls or extract required fields. Error handling is minimal; no recovery guidance. Composition is sound (single responsibility per tool), but include/receipt patterns add complexity without clear benefit.
Get effective gates for a project
List tasks in a project
Acknowledge a signal
Poll for signals (inbox items)
Release a claim without finishing
Create a new task
Finish a task — work-claim or review-claim
Output schemas not documented. LLMs cannot infer what fields are returned, breaking downstream tool chaining and forcing discovery calls.
Descriptions lack WHEN/WHY guidance. 'Pick up the next signal' does not explain when to call task_pickup vs tasks_get, or what 'signal' means in domain context.
Complex include/receipt patterns (include arrays with enum options) add cognitive load without clear benefit. Simplify to fixed response shapes or use separate tools.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 64 | 2026-07-28+ | v2 |
Add a note to a task
Pick up the next signal, review, or work item
Update task description and/or template data
Begin work on a task (atomic claim + transition)
Attach branch + PR metadata to a work-claimed task
Get a single task by ID
Get workflow primer documentation
No error handling guidance. Tools do not document what errors are retryable, user-fixable, or fatal, or how to recover.
Parameter descriptions lack format/constraint details. 'Task ID (UUID)' is present but no guidance on what happens if ID is invalid or not found.