Simple reminder system for Claude - like 'Alexa, remind me...' but for AI. Leave notes for yourself across sessions, check them when you start, act on them or save them.
The server provides 7 tools with explicit schemas and descriptions, but quality is inconsistent. All tools are properly registered with JSON Schema input definitions and non-empty descriptions. However, descriptions are brief (10-40 chars), lack WHEN-to-use context, and omit output schema documentation. Parameter descriptions are minimal or missing. Error handling lacks recovery guidance. No destructive operation confirmations despite delete/clear tools. Tool names follow verb_noun convention well. Schemas are present and typed, but parameter constraints are minimal (only priority and filter use enums). No validation rules documented. Output structure is not formally specified for any tool.
Check your reminders
Clear old completed/moved reminders
Mark a reminder as completed
Delete a reminder
Get help on using reminders
Move reminder to permanent notes
Add a reminder
Descriptions are critically short (10-50 chars). No WHEN-to-use context or actionable guidance for LLM tool selection.
No output schemas documented. LLMs cannot predict return structure (e.g., what fields are in check_reminders response? What does remind_me return?). This breaks chaining and forces exploratory calls.
Destructive operations (delete_reminder, clear_old_reminders) lack dry-run or confirmation pattern. Agents can permanently delete reminders without reversibility.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 8 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 39 | - | v1 |
Parameter 'filter' in check_reminders and 'days' in clear_old_reminders have defaults documented in description text, not JSON Schema. Schema lacks minimum/maximum bounds or default declarations.
Error responses are not shown in handler code. No recovery guidance (e.g., if delete_reminder fails with ID not found, what should the LLM do next?). No error categorization (retryable vs. fatal).
Parameter 'id' in delete_reminder, complete_reminder, move_to_notes lacks format specification. No guidance on ID format (is it 'rem_...'? Can a user pass a partial string? Should fuzzy matching be supported?)
move_to_notes writes to hardcoded Obsidian path (~/Documents/Obsidian/Brain/Reminders). This assumes a specific local filesystem structure and will silently fail if the path doesn't exist. No configurable vault support.
HELP tool description is vague ('Get help on using reminders'). This is a meta-tool that provides documentation, its purpose is clear from the name alone, making the description redundant. Better: use for discovery of available operations.