Backend-agnostic AI photography workflow agent (v0.3 alpha).
Photo-agent has 6 tools with basic schemas and descriptions, but significant gaps prevent higher scores. All tools have input schemas with type definitions and descriptions, meeting minimum requirements. However, descriptions are generic and lack LLM-optimized guidance on WHEN to use each tool or what distinguishes them. Parameter descriptions are minimal (e.g., 'Path to the photo file' repeated across tools). No output schemas are documented. Error handling is absent, tools provide no recovery guidance. Tool names follow verb_noun convention (get_, set_, create_, export_) but lack clarity on distinctions (e.g., create_develop_preset vs create_virtual_copy both create checkpoints but serve different workflows). No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk classifications in metadata.
Create a checkpoint/preset from current develop settings (Lightroom MCP tool operation)
Create a workflow copy (virtual copy) of a photo (Lightroom MCP tool operation)
Render preview/export of edited photo (Lightroom MCP tool operation)
Read current edit state from the backend (Lightroom MCP tool operation)
Reconcile workflow copy back to master (Lightroom MCP tool operation)
Apply global adjustment settings to a photo (Lightroom MCP tool operation)
No output schemas documented. LLMs cannot infer what fields are returned or plan downstream calls. All 6 tools lack return type documentation.
Descriptions are generic and under 50 characters. 'Read current edit state from the backend' and 'Apply global adjustment settings to a photo' lack context on WHEN to use each tool or what distinguishes similar operations (preset vs virtual copy).
No error handling or recovery guidance. Tools provide no actionable error messages, retryability classification, or next steps if a call fails.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 51 | 2025-06-18+ | v2 |
No tool annotations despite clear risk metadata. Tools marked WRITE (5 tools) lack destructiveHint annotation. get_photo_metadata lacks readOnlyHint. Annotations enable LLM safety reasoning.
Parameter descriptions are minimal and repetitive. 'Path to the photo file' appears 5 times without explaining format, constraints, or expected file types. 'Develop settings to apply' lacks structure guidance.