Desktop interaction hub for AI agents: notifications, blocking questions and sounds that cannot be missed the way a terminal can
Both tools have minimal descriptions (under 100 chars) and lack parameter descriptions entirely. Schemas are present but sparse. park_quiet and park_progress are testing/diagnostic tools with no real business logic, making them unsuitable for production agent use. Naming is clear (verb_noun pattern), but parameter documentation is absent. No error handling guidance, no output schema documentation, and no indication of what these tools return beyond implicit behavior.
Park for `seconds`, sending a progress notification every `every` seconds, then return.
Park for `seconds` in complete silence, then return. Used to measure how long this client tolerates a silent tool call.
Parameter descriptions missing entirely. 'seconds' parameter in both tools has no description explaining units, constraints, or valid range.
Output schema not documented. Tools return implicit results with no documented structure, field names, or types for LLM consumption.
Tool descriptions are under 100 characters and lack WHEN/WHY context. LLMs cannot determine when to select these tools over alternatives.
No error handling guidance. No documentation of what happens if seconds is negative, zero, or exceeds system limits.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 44 | 2026-07-28+ | v2 |
park_progress parameter 'every' lacks description. Unclear if it must be less than 'seconds', what happens if equal, or minimum/maximum bounds.