Content
68%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.
A concise, largely executable operations cookbook with clear sectioning and useful cross-references to sibling skills. Its main weaknesses are the broken shell-expansion inside the quoted Python heredocs and the absence of validation steps for the batch operations (split, merge, render-all).
Suggestions
Fix the quoted-heredoc expansion bug: inside <<'PY' the $(basename "$PWD") construct is literal, so pypdf writes to a nonexistent directory. Use plain filenames ('input.pdf', 'merged.pdf') or pass the path via sys.argv/environment variable instead.
Add validation after batch operations, e.g. verify a split produced the expected file count (ls page_*.pdf | wc -l vs pdfinfo's Pages) or confirm pdfinfo succeeds on the merged output before reporting success.
Replace the ten repetitions of /workspace/$(basename "$PWD")/ with a one-line note about the workspace path convention, which would trim token cost and remove the confusing construct from the bash examples too.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is a lean recipe book with no explanations of concepts Claude already knows and a genuinely useful tool inventory, but the opaque `/workspace/$(basename "$PWD")/` prefix is repeated in every command where a single path note would do — minor trimmable noise. | 4 / 5 |
Actionability | Bash commands (pdftotext, pdfinfo, pdftoppm) are concrete and executable, but the Python heredocs use a quoted delimiter (<<'PY') so `$(basename "$PWD")` inside them is never expanded by the shell, producing literal broken output paths; the merge example also mixes relative input paths with that absolute construct. Mostly executable with one systematic defect. | 4 / 5 |
Workflow Clarity | Each single-task recipe is unambiguous, but split, merge, and render-all are batch operations and the skill includes no validation or verification steps (e.g., checking the split output's page count), which caps workflow clarity at 3 per the rubric. | 3 / 5 |
Progressive Disclosure | The ~80-line body is cleanly sectioned per operation with clear cross-skill pointers and nothing inlined that belongs in a separate file; it exceeds the 50-line simple-skill threshold and offers no quick-start/advanced split or reference structure, so it sits at good-but-not-ideal organization. | 4 / 5 |
Total | 15 / 20 Passed |