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 action-oriented and executable with concrete commands and examples, but it is substantially padded with boilerplate, has no validation feedback loop in the batch-screening flow, and references a missing PDF, holding it near the midpoint.
Suggestions
Remove generic template boilerplate (Security Checklist, Evaluation Criteria, Lifecycle Status, Output Requirements, Response Template, User Checkpoints) that does not add skill-specific knowledge.
Add an explicit validate->fix->retry checkpoint to the screening workflow (e.g. run `python -m py_compile scripts/main.py`, confirm expected output files, re-run on parse failure) before declaring results.
Either add the missing references/prisma_2020_checklist.pdf or remove it from the References list so signaled paths resolve.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The 348-line body is noticeably padded with generic template boilerplate (Security Checklist, Evaluation Criteria, Lifecycle Status, Risk Assessment, Output Requirements/Contract, Response Template, etc.) that restates obvious practices rather than adding domain knowledge Claude lacks, and it duplicates the identical `python scripts/main.py --help` command twice in the audit-ready block. | 2 / 5 |
Actionability | It provides mostly executable guidance — concrete CLI invocations (`python scripts/main.py --input references.csv --criteria criteria.yaml --prisma`), a complete criteria YAML example, an output-file table, and a PRISMA JSON shape — with only minor gaps (e.g. `pip install -r requirements.txt` with no bundled requirements file, and duplicated help commands). | 4 / 5 |
Workflow Clarity | A Workflow and Decision Algorithm are sequenced and there is a fallback path, but the batch/destructive screening operations lack an explicit validate-then-fix feedback loop with concrete checkpoints in the main flow; the rubric caps batch-operation skills without clear validation feedback loops at 3. | 3 / 5 |
Progressive Disclosure | Structure is present with a Quick Check, References section pointing to criteria_template.yaml and sample_references.csv (both real), but one referenced file (references/prisma_2020_checklist.pdf) does not exist and navigation/signaling is inconsistent, leaving organization at the 'some structure but could be better organized' anchor. | 3 / 5 |
Total | 12 / 20 Passed |