Local-first Windows project progress control plane for AI agents. Provides task management, progress tracking, and engineering metrics via MCP protocol with core, memory, and actions profiles.
Tool definitions lack visible input schemas and parameter descriptions in the provided source. Registry file shows tool names and brief descriptions (10-60 chars), but no JSON Schema definitions, parameter types, or constraints are evident. Tool names follow verb_noun convention (atm_begin, atm_task_list, atm_task_get, atm_task_patch), which is positive. However, descriptions are minimal (e.g., 'List tasks in the current project with optional filtering' for atm_task_list). No parameter documentation visible. Error handling and output schemas not documented in source. Composition is reasonable (9 focused tools), but lack of schema visibility prevents confident assessment of parameter quality, validation, and output structure.
Start a new task session. Returns a brief summary of current working set.
Get a brief summary of the current working set. Only call after context compression or long absence.
End the current task session and release all claimed tasks.
Get engineering metrics for a project or work item including git-based change tracking.
Add progress to a task with scope and percentage.
Record progress or notes on a task.
Get detailed information about a specific task.
No input schemas visible in source code. Cannot verify parameter types, constraints, enums, or validation rules for any of the 9 tools.
Tool descriptions are minimal (10-60 chars), below the 50-200 char baseline for LLM-optimized descriptions. Descriptions lack WHEN to use, prerequisites, and expected output structure.
No parameter descriptions visible. LLMs cannot infer meaning of parameters like 'scope', 'percentage', 'operation_type' from names alone.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 39 | 2025-06-18+ | v2 |
List tasks in the current project with optional filtering.
Apply normalized task operations (16 standardized operation types).
Output schemas not documented. Unclear what fields atm_task_list, atm_task_get, and atm_engineering_metrics return, preventing LLMs from planning downstream tool calls.
atm_task_patch accepts '16 standardized operation types' but no enum or documentation of valid operations visible. LLMs will hallucinate invalid operation names.