Content
67%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.
This is a strong, highly actionable workflow skill: concrete commands, venue-specific templates, explicit anti-hallucination gates, and genuine validation loops (reviewer-model pass, reverse outline, final checklist). Its weaknesses are redundancy — reference pointers and several rules are repeated across sections — and dependency on unverifiable sibling-skill paths for its templates and writing guides.
Suggestions
Consolidate the three sections that point to ../ml-paper-writing references (Orchestra-Guided Writing Overlay, Step 5 pointer, and Writing Quality Reference) into one, and state each rule (natbib-vs-cite, IEEE page limits, backup) once in Key Rules only.
Replace the comment-only Python bib-filtering block with a runnable script, and add an explicit final step to compile (e.g., `latexmk -pdf main.tex`) and fix errors until the build is clean, closing the workflow's feedback loop.
Vendor the referenced writing-guide/checklist/citation-workflow material into this skill's own references/ directory (or ship the templates alongside it) so the skill is self-contained and its referenced paths are verifiable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient, instruction-dense prose with little explanation of concepts Claude already knows, but it could be tightened: the ../ml-paper-writing reference pointers are repeated across three separate sections (Orchestra-Guided Writing Overlay, Step 5, and Writing Quality Reference), and the natbib-vs-cite rule, IEEE page-limit rule, and backup rule are each restated 2-3 times (Step 3/Step 4/Key Rules/Final Checks). This matches anchor 3's "some unnecessary explanation or could be tightened"; not 4 because the cross-section repetition is more than minor. | 3 / 5 |
Actionability | Largely executable guidance: exact curl commands for DBLP and CrossRef, copy-paste LaTeX preambles per venue, a concrete file tree, an exact MCP codex call with model and reasoning-effort config, and word/page targets per section. Not 5 because the Python bib-filtering block is comment-only pseudocode ("import re" followed by # steps) rather than runnable code, and the compile expectation names latexmk without giving the command. | 4 / 5 |
Workflow Clarity | Steps 0-8 are clearly sequenced with real checkpoints: backup-before-overwrite of paper/, the verified BibTeX fallback chain ending in a [VERIFY] no-fabrication gate, a reviewer-model pass with severity levels and "Apply CRITICAL and MAJOR fixes", a reverse-outline test, and a final checklist. Not 5 because there is no explicit compile-and-fix-errors feedback loop — the skill requires latexmk-clean output and checks \ref/\label matching but never instructs running latexmk and iterating on errors. | 4 / 5 |
Progressive Disclosure | Structure is good: well-organized sections with one-level-deep, purpose-conditioned references ("Read … before drafting the Abstract", "only when the built-in DBLP/CrossRef workflow is insufficient"). Not 5 because no bundle files exist in this skill's directory — all references point into a sibling skill (../ml-paper-writing/) that is not part of this bundle and cannot be verified — and a fair amount of detail (venue preamble snippets, per-section writing guidelines, citation rules) is inlined in the ~370-line body rather than split out. | 4 / 5 |
Total | 15 / 20 Passed |