Content
81%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A dense, highly actionable reference with exemplary validation and error-recovery workflows for batch uploads and versioning. Its main weakness is redundancy: the upload-method selection logic is stated three times and several paragraphs could be tightened, and the sheer inlined detail keeps it from being a lean overview with split-out references.
Suggestions
State the upload-method selection logic (host-assisted vs signed-URL vs CLI) once — e.g., in a decision table in the "MCP Image and Video Upload" section — and have the Methods table and "Video Upload" section refer to it instead of restating the conditions.
Break the long video-polling and annotation-group paragraphs into short bulleted rules (poll each videoId, use final videoId, report failed-chunk counts as unverified) so they are scannable rather than run-on prose.
Move the upload edge-case detail (screening verdicts, dedup ID handling, chunk retry semantics) into a reference file like upload-details.md linked from the main body, keeping SKILL.md as a tighter overview.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Tables are token-efficient and nearly every sentence carries unique operational facts, but the upload-method selection guidance and signed-URL fallback conditions are repeated across the Methods table, "Video Upload", and "MCP Image and Video Upload" sections, and paragraphs like the video-polling and annotation-group ones are long run-ons that could be tightened. | 3 / 5 |
Actionability | Fully executable guidance throughout: copy-paste CLI and SDK snippets ("roboflow import -w <workspace> -p <project-id> /path/to/dataset", "project.version(1).download(\"yolov8\")"), exact limits ("Max 20 MB per image", "up to 100 files / 1 GB per call"), concrete RoboQL examples ("class:helmet AND NOT (tag:v1 OR tag:v2)"), and specific tool names and status values covering the common cases. | 5 / 5 |
Workflow Clarity | Multi-step batch upload flows carry explicit validation and feedback loops: poll "image_upload_status" until "completed" or "failed", per-video "video_upload_status" polling, a safe chunk-failure retry procedure with expected duplicate errors, screening verdict handling, and a numbered version-creation pipeline; irreversible class changes carry a warning with a non-destructive alternative. | 5 / 5 |
Progressive Disclosure | Clear section organization with a well-signaled, one-level-deep "Related Pages" section (labeling.md, cloud-storage SKILL.md), but at ~250 lines the body reads as a full reference rather than an overview — the upload edge-case deep-dives (screening, dedup IDs, chunk retries) could be split into a reference file. | 4 / 5 |
Total | 17 / 20 Passed |