Content
63%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 well-structured, largely actionable skill with a clear decision tree and thoughtful tracked-changes guidance (batching, minimal-edit principle, final verification). Its weaknesses are duplicated batching/boilerplate text, deferred core API examples, no pre-pack XML validation loop, and references that point to files absent from the provided bundle while the real scripts/ files go undocumented.
Suggestions
Deduplicate the batch-grouping guidance (step 2 'Batch organization' vs step 4 'Suggested batch groupings') and the thrice-repeated 'MANDATORY - READ ENTIRE FILE' boilerplate — state each once.
Fix reference paths to match the actual bundle: docx-js.md, ooxml.md, and ooxml/scripts/{unpack,pack}.py do not exist; point to the real scripts/ layout (document.py, utilities.py, templates/) or add the referenced files.
Add an explicit XML validation checkpoint after each batch (e.g., a validate step on the unpacked directory) with a fix-and-retry loop before running pack.py, instead of relying solely on post-pack markdown verification.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly lean (commands and examples over concept explanations), but carries real duplication: the batch-grouping guidance appears nearly verbatim twice ("**Batch organization** (group 3-10 related changes per batch)" in step 2 and "**Suggested batch groupings:**" in step 4), and the "MANDATORY - READ ENTIRE FILE ... NEVER set any range limits" boilerplate is repeated three times. This matches 'mostly efficient but includes some unnecessary explanation or could be tightened' rather than level 4's 'minor instances'. | 3 / 5 |
Actionability | Mostly executable: concrete pandoc/soffice/pdftoppm commands with flags, unpack/pack invocations, a grep-based verification snippet, and a GOOD/BAD XML example for minimal edits. It falls short of level 5 because the core editing mechanics (Document library methods, get_node usage, docx-js Document/Paragraph/TextRun syntax) have no inline example and are deferred entirely to ooxml.md / docx-js.md, and the tracked-changes snippet is a fragment rather than a runnable script. | 4 / 5 |
Workflow Clarity | A decision tree routes tasks to clearly numbered workflows, and the redlining workflow includes per-batch testing ("Test each batch before moving to the next"), a re-grep-before-each-script note, and a final markdown verification with grep checks. It stops short of level 5 because there is no XML-level validation checkpoint before packing and no explicit fix-and-retry feedback loop — verification is post-hoc output checking only, which fits 'clear sequence with most checkpoints present; minor validation gaps'. | 4 / 5 |
Progressive Disclosure | Sectioning and one-level-deep references are well signaled ([docx-js.md], [ooxml.md]), but scored against the actual bundle the references are broken: docx-js.md, ooxml.md, and ooxml/scripts/unpack.py|pack.py do not exist, while the files that do exist (scripts/document.py, scripts/utilities.py, scripts/templates/) are never referenced or made discoverable from the body. This is more than the level-4 'minor organization gaps' — navigation from SKILL.md to the real bundle fails, matching level 3's 'references present but' mis-organized structure. | 3 / 5 |
Total | 14 / 20 Passed |