Content
50%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.
The body is rich in executable, Domino-specific guidance but is padded with redundant ASCII trees, frontmatter-duplicating sections, and generic data-engineering tutorials. It would score much higher if split into reference files with a lean overview and trimmed to what Claude doesn't already know.
Suggestions
Delete the '## Description' and '## Activation' sections — they duplicate the frontmatter — and drop generic tutorials (pandas chunking, Dask, numpy.memmap, parquet/feather/HDF5 trade-offs) that Claude already knows, keeping only the Domino-specific mount paths.
Collapse each ~60-line ASCII directory tree into the compact path table that already follows it, keeping one small illustrative tree at most.
Split bulk material into one-level-deep reference files (e.g., PATHS.md for mount structures, BEST-PRACTICES.md for organization/format guidance, TROUBLESHOOTING.md) and reference them from a lean SKILL.md overview.
Add validation checkpoints to the snapshot and upload workflows, e.g., verify the snapshot appears under /snapshots/{dataset}/{tag} before modifying data, and confirm uploaded file counts/sizes before proceeding.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Noticeably verbose: '## Description' and '## Activation' duplicate the frontmatter; two ~60-line ASCII directory trees are largely duplicated by the path tables directly beneath them; and sections like 'Reading Large Datasets' teach generic pandas/Dask/numpy patterns Claude already knows rather than Domino-specific guidance. | 2 / 5 |
Actionability | Mostly executable: concrete SDK calls (domino.datasets_create, datasets_snapshot, datasets_tag), a runnable project-type detection snippet, specific mount paths, and CLI upload commands. Minor gaps — the metadata example uses json.dump without importing json, and the '@tag' snapshot path syntax and domino datasets_tag parameters are unverified. | 4 / 5 |
Workflow Clarity | Sections are sequenced by task and the project-type check is a genuine checkpoint, but the snapshot and upload workflows lack validation steps (e.g., verify a snapshot was created before modifying data, confirm files landed after upload) — checkpoints are mostly implicit, matching the 'validation gaps' anchor. | 3 / 5 |
Progressive Disclosure | A 390-line monolithic file, well beyond the 50-line simple-skill exception. Bulk content that belongs in separate reference files (the two directory trees, best practices, large-dataset reading, troubleshooting) is fully inlined; the only references are external documentation URLs, with no internal bundle structure to navigate. | 3 / 5 |
Total | 12 / 20 Passed |