directory-pipeline

Builds the exact `python main.py` command for the directory-pipeline project. Trigger this skill when the user asks "what command should I run", "how do I run the pipeline", "what flags do I need", "run the pipeline", or describes a goal like "extract entries" or "add bounding boxes". The skill inspects the output directory to detect what has already been run, then outputs a precise, ready-to-paste command with an explanation of what each flag does.

Diagnoses problems with the directory-pipeline. Trigger this skill when the user reports something isn't working: "no thumbnails", "empty CSV", "no bounding boxes", "canvas_fragment has no xywh", "entries are missing", "pipeline failed", or any symptom suggesting a stage produced wrong or missing output. The skill asks for the symptom, runs targeted file checks, and gives the exact fix command.

Builds the command to launch the interactive data explorer for the directory-pipeline. Trigger this skill when the user asks "how do I explore the data", "open the explorer", "view the entries", "create the HTML explorer", "browse the CSV", or mentions `explore_entries.py`. The skill inspects the output directory to detect available files and builds the right command with the correct flags for thumbnails, palette, and translated variants.

Walks the user through first-time calibration for a new collection type in the directory-pipeline. Trigger this skill when the user says "I have a new collection", "starting fresh", "never run this before", "how do I calibrate", "first time setup", or asks about `--select-pages` or `--generate-prompts`. Explains the "calibrate once, run many" pattern and walks through the three setup steps in sequence.

Conducts a structured expert UX/HCI review of a data-browser explorer HTML file. Trigger when the user asks for a "UX review", "HCI review", "usability review", "interface feedback", or "how can I improve the explorer". The skill guides a multi-pass review combining source-code analysis with user-supplied screenshots or observations, and produces prioritised, actionable recommendations.