Content
77%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.
An exceptionally actionable and well-sequenced body with validation and feedback loops throughout — the strongest possible guidance for this tool. Its weaknesses are verbosity (repeated caveats and one section re-teaching a known concept) and a monolithic single-file layout that inlines material (XML appendix, specialty recipes) better suited to separate reference files.
Suggestions
Split the raw-set XML appendix, the advanced/specialty topics, and the report-level recipes into references/ files (e.g., references/raw-set-xml.md, references/recipes.md) and link them one level deep, shrinking SKILL.md to the overview, workflow, and QA gates.
Deduplicate recurring caveats: state the unsupported `differentFirstPage` prop and the `updateFields=true` guidance once each (in Common Pitfalls and TOC respectively) and reference them elsewhere instead of re-explaining them in 3-4 sections.
Trim the 'Mental Model' section to only what is novel (the semantic-path API over the ZIP) and drop the re-explanation that a .docx is a ZIP of XML parts, which Claude already knows.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly tool-specific knowledge Claude would not already know, but with tightening opportunities: the unsupported `differentFirstPage` warning is repeated in three sections, `updateFields=true` is re-explained in 4+ places, and the 'Mental Model' section explains that a .docx is a ZIP of XML parts — a concept Claude already knows. This fits the anchor 'mostly efficient but includes some unnecessary explanation or could be tightened'; it is not a 4 because the repetition is more than minor. | 3 / 5 |
Actionability | Every section ships executable, copy-paste bash — concrete `add`/`set`/`query` commands with real prop values, a fully scripted Delivery Gate with reject/exit logic, and recipes flagged as executed and `validate`-passed. This matches the anchor for fully executable, copy-paste-ready commands covering the common cases. | 5 / 5 |
Workflow Clarity | A six-step Common Workflow with explicit sequencing, incremental execution with exit-code checks, a QA minimum cycle, a Delivery Gate with pass/REJECT branches, and a fix-then-rerun-full-cycle feedback loop — exactly the anchor's 'clear sequence with explicit validation steps; feedback loops for error recovery' for batch/XML manipulation. Not a 4 because validation checkpoints and recovery loops are pervasive rather than mostly present. | 5 / 5 |
Progressive Disclosure | The skill is a single ~560-line file with no references/, scripts/, or assets/ bundle at all; content that clearly belongs in separate files (the raw-set XML appendix, the advanced specialty topics, the report-level recipes) is inlined. Section signposting is good ('skip if you are writing a report', delegating full schemas to `officecli help`), which lifts it above a 2, but there are no one-level-deep reference files, so it cannot reach the 4 anchor's 'most content appropriately placed' across files. | 3 / 5 |
Total | 16 / 20 Passed |