Content
71%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 well structured with a decision tree, concrete commands, and a properly validated redlining workflow, but it is undermined by references that do not resolve in the bundle and by triplicated batching guidance. The delegation pattern is sound; the referenced files and paths simply are not shipped alongside it.
Suggestions
Consolidate the batching guidance: the 'Batching Strategy' paragraph, step 2's 'Batch organization', and step 4's 'Suggested batch groupings' repeat the same section/type/proximity advice three times — state it once and reference it from the other spots.
Fix reference paths or ship the referenced files: docx-js.md, ooxml.md, and ooxml/scripts/unpack.py|pack.py do not exist in the bundle (scripts live at scripts/ with no unpack/pack scripts), so every workflow's first steps point at missing targets.
Add an output verification step to the create-new-document and basic-editing workflows (e.g., pandoc-convert the packed .docx and inspect it), mirroring the redlining workflow's final verification so all workflows end with a validation checkpoint.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and actionable (decision tree, copy-paste commands, a GOOD/BAD XML example), but the batching guidance is repeated three times — the "Batching Strategy" paragraph, step 2's "Batch organization", and step 4's "Suggested batch groupings" all restate the same by-section/by-type/by-proximity advice, which could be trimmed. Not anchor 5 because this duplication means not every token earns its place; not anchor 3 because the over-explanation is localized rather than pervasive. | 4 / 5 |
Actionability | Concrete executable commands throughout — pandoc with --track-changes options, unpack/pack invocations, soffice and pdftoppm with flag explanations, and grep-based verification — plus a concrete GOOD/BAD tracked-changes XML example. However, the central scripting step of both edit workflows is delegated entirely to ooxml.md and docx-js.md, which are not present in the provided bundle, leaving a gap; not anchor 5 because the workflows cannot actually be executed end-to-end from what ships here. | 4 / 5 |
Workflow Clarity | The redlining workflow has a clear numbered sequence with explicit validation checkpoints (grep before each script, test each batch, and a final verification step with pandoc conversion and grep checks), satisfying the batch-operation cap. Not anchor 5 because the create-new-document and basic-editing workflows end at export/save with no output verification or error-recovery loop, and error-recovery guidance (what to do when a batch fails) is only implied. | 4 / 5 |
Progressive Disclosure | The design intent is good — an overview with a decision tree delegating detail to well-signaled one-level-deep references with mandatory-read instructions and line counts. But scored against the actual bundle, none of the referenced targets resolve: docx-js.md, ooxml.md, and ooxml/scripts/unpack.py|pack.py are absent (scripts live under scripts/ and contain no unpack/pack), so the navigation paths the body depends on are broken. | 3 / 5 |
Total | 15 / 20 Passed |