Content
50%Weight 40%Scale 1-3Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is a well-organized, mostly lean CLI reference catalog that gives real commands and flags. It scores uniformly at 2 because it lacks concrete worked examples, validation steps for destructive operations, and any file-based progressive disclosure, while not being verbose or disorganized enough to drop to 1.
Suggestions
Add a few concrete worked examples with real values, e.g. `hf download bert-base-uncased --include "*.json" --local-dir ./bert`, instead of placeholder-only signatures.
Add validation/verification guidance for destructive and batch operations, e.g. run `--dry-run` first or verify with `hf repos info`/`hf buckets info` after delete/remove operations.
Reduce inline flag duplication by keeping only the most-used flags in SKILL.md and pointing the full catalog to `hf <command> --help`, or move the full command reference to a one-level-deep REFERENCE.md.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is a lean command catalog with no concept explanations Claude already knows, but ~170 lines list every flag inline across commands (e.g. repeated --quiet, --namespace) while the Tips section admits `hf <command> --help` provides full options, so it could be tightened; not score 3 because the flag listings duplicate what --help already offers. | 2 / 3 |
Actionability | Real executable commands and a concrete install line are given, but signatures use placeholders (REPO_ID, NAME) with no worked examples using real values, and real-world usage is deferred to `--help`; not score 1 because the commands are concrete and runnable, not abstract. | 2 / 3 |
Workflow Clarity | Commands are well organized by group but there is no sequenced multi-step workflow, and destructive/batch operations like `hf repos delete` and `hf buckets remove` lack validation or verification steps, capping this at 2 per the rubric; not score 1 because install→auth→use is implied and commands are clearly structured. | 2 / 3 |
Progressive Disclosure | Sections are well organized by command group, but the entire reference is a monolithic inline block (well over 50 lines) with no one-level-deep references to split out detail; the simple-skill carve-out does not apply, so it is not score 3, though there is no nested-reference mess keeping it above 1. | 2 / 3 |
Total | 8 / 12 Passed |