Content
48%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 task-specific core (the two-phase screening workflow with a JSON output contract and script validation) is clear and well sequenced, but it is embedded in ~80 lines of generic template filler. More seriously, the screening logic and schema are delegated to references/screening_prompts.md and scripts/screen_paper.py, neither of which exists in the bundle, leaving the skill's central details unresolvable.
Suggestions
Ship or fix the missing bundle files: references/screening_prompts.md (screening logic + JSON schema) and scripts/screen_paper.py are cited as the primary implementation surface but do not exist in references/ or scripts/, so the skill's core logic is inaccessible.
Delete the generic template sections (When to Use, Key Features, Implementation Details, Required Inputs, Output Contract, Validation and Safety Rules, Failure Handling, Input Validation, Quick Validation) — they restate the description or repeat boilerplate that Claude already knows; keep the Workflow and Resources sections.
Remove the leaked generation path in Example Usage ("cd 20260316/scientific-skills/Data Analytics/meta-abstract-screener") and replace it with a path relative to the skill directory; also include an example screening input/output pair inline so the skill is actionable even before opening the references.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Noticeably verbose with several padded sections: "Key Features" restates the frontmatter description verbatim, "When to Use" offers only circular filler ("Use this skill when the request matches its documented task boundary"), and "Implementation Details", "Required Inputs", "Output Contract", "Validation and Safety Rules", "Failure Handling", and "Input Validation" are generic template boilerplate that adds nothing specific to abstract screening. It avoids level 1 because it does not explain concepts Claude already knows (no PDF-style tutorials), but roughly half the body is unnecessary template text that could be cut. | 2 / 5 |
Actionability | There are concrete commands ("python scripts/screen_paper.py '<json_output>'", "python scripts/screen_paper.py --help", py_compile check) and a concrete output contract ("The output must contain strictly `Result` and `Reason`"), but the actual decision logic and JSON schema are delegated to "references/screening_prompts.md" (Steps 1 and 2), which does not exist in the bundle — the referenced files are missing, so the key executable details are inaccessible. This is "some concrete guidance but incomplete; missing key details" rather than mostly executable guidance. | 3 / 5 |
Workflow Clarity | The core Workflow is a clearly sequenced three-phase process (Analysis → Formatting → Optional Validation via the script), with an additional Quick Validation path and a Failure Handling section covering error recovery in generic terms. It falls short of level 5 because validation is only optional and the failure-recovery guidance is boilerplate rather than concrete feedback loops tied to the screening steps; it is above level 3 because checkpoints are explicitly present, not merely implied. | 4 / 5 |
Progressive Disclosure | Structure is decent — a Resources section clearly signals one-level-deep references ("references/screening_prompts.md", "scripts/screen_paper.py") — but the bundle contains no references/ or scripts/ directories, so every detailed reference points to a nonexistent file and the promised deeper content is unavailable. Combined with substantial inline generic boilerplate that should be trimmed, this sits between "references present but not clearly signaled" (they are signaled, but broken) and the well-placed level-4 anchor. | 3 / 5 |
Total | 12 / 20 Passed |