Content
56%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 skill's actual core content — a two-step Vision-LLM-then-script workflow with concrete formatting rules and an example — is well specified and executable. However, the body is buried under extensive generic boilerplate, duplicated validation sections, and cross-references that add tokens without skill-specific value, and the documented commands do not fully match the packaged script's real CLI.
Suggestions
Remove the generic template sections (When to Use, Key Features, Implementation Details, Deterministic Output Rules, Completion Checklist, etc.) that contain no skill-specific information, keeping only the Workflow, Usage, Examples, and validation guidance.
Merge the duplicate "Validation Shortcut" and "Quick Validation" sections, and replace the circular "See `## Usage` above" cross-references with the actual content placed once.
Document the real CLI of `scripts/format_result.py` (positional `text language` arguments) and the Chinese header/footer formatting branch, and remove the fabricated expected-output block.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Roughly two-thirds of the body is generic template boilerplate with no skill-specific content ("Use this skill when the request matches its documented task boundary", "keep results reproducible, identify assumptions explicitly", and ~10 similar sections). It is not 1 because it never explains concepts Claude already knows and the core Workflow/Usage/Examples sections are efficient; it is not 3 because the padding is extensive, duplicated across sections, and includes a time-sensitive date-stamped path ("cd \"20260316/scientific-skills/...\""). | 2 / 5 |
Actionability | The Workflow gives precise formatting rules ("Inserts `(Figure 2)` before the last punctuation mark", exact header/footer strings) plus a worked input→process→output example and runnable commands. It is not 5 because the documented validation command `python scripts/format_result.py --help` does not match the script's actual positional `text language` CLI, and the script's Chinese header/footer branch is undocumented in the body. | 4 / 5 |
Workflow Clarity | The two-step sequence (Vision LLM image analysis with explicit prompt guidelines → script formatting) is unambiguous, with validation shortcuts, a completion checklist, and failure handling. It is not 5 because the "Expected output format" block after the validation command is aspirational rather than what the script produces, and the duplicated validation sections create minor confusion. | 4 / 5 |
Progressive Disclosure | Both referenced bundle paths (`scripts/format_result.py`, `scripts/validate_skill.py`) exist and are one level deep, but the SKILL.md organization is muddled: "Validation Shortcut" and "Quick Validation" duplicate each other, sections point forward and backward with "See `## Usage` above", and boilerplate sections are interleaved with the real content. It is not 4 because these duplications and circular cross-references are more than minor organization gaps; it is not 2 because the file does have clear section headers and all references resolve. | 3 / 5 |
Total | 13 / 20 Passed |