Content
60%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 a well-referenced, well-sequenced orchestrator document with genuine validation checkpoints and read-back loops, and all linked playbooks exist. Its main weaknesses are verbosity — dense, repetitive policy prose that buries the operational signal — and a bundle inconsistency (a required update script with no scripts/ directory present).
Suggestions
Trim the repeated prohibitions and policy digressions ("Never..." clauses appear in nearly every section) into a single boundary section or the referenced playbooks; the body could be roughly halved without losing operational content.
Resolve the bundle inconsistency: either ship scripts/paperspine_update.py (and the author_voice_check.py / humanize_check.py helpers) or reword the update and quality-check steps so they do not invoke files absent from the skill.
Inline the exact call sequence for the launch/configure/open-task loop (or a minimal example of one paperspine_open_task payload) since the body currently says "Follow the exact calls in product-v1-workflow.md" for its most safety-critical step.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is ~280 lines of dense, policy-laden prose with heavy padding: prohibitions are repeated many times ("Never invent...", "Never fake review/delivery completion", "Never repeatedly reinstall"), and abstract policy statements like "The manuscript is the product; the task state is only its support" compete with operational content. It does not explain concepts Claude already knows (so not level 1), but several sections are noticeably padded and could be cut or pushed to the referenced playbooks. | 2 / 5 |
Actionability | Concrete anchors are present: "scripts/paperspine_update.py --preflight --yes", "launch --no-open", tool names ("paperspine_open_task", "paperspine_commit_milestone", "paperspine_authorize_materials") with exact field names ("grants[].include_paths", "skill_bridge.web_path") and concrete success conditions ("source=web_user", "user_confirmed=true", "readiness.ready=true"). It is not level 5 because most tool-call syntax is deferred to product-v1-workflow.md and the referenced update script is not present in this bundle (no scripts/ directory), leaving a few gaps. | 4 / 5 |
Workflow Clarity | A clear 7-section scientific workflow is sequenced (anchor → research → contribution → writing/figures → render → review → delivery) with explicit validation checkpoints: "Only then begin the next dependent segment", mandatory read-back after "paperspine_commit_milestone", "Inspect every PDF page and the actual DOCX", and feedback loops ("If read-back fails, fix the concrete binding/sync error before claiming progression"; "After two unchanged attempts, change tactic"). Not level 5 because the density and interleaved policy digressions make several checkpoints implicit and the exact call sequence is delegated to references rather than stated inline. | 4 / 5 |
Progressive Disclosure | Structure is good: every section signals one-level-deep markdown links to real files in references/ (all 34 referenced files verified to exist), organized by workflow stage with clear "Read X as applicable" navigation. It is not level 5 because the SKILL.md itself inlines substantial operational detail (update authority, Web-interruption policy, milestone mechanics) that belongs in the referenced playbooks, and it cites "scripts/paperspine_update.py" which is absent from the bundle. | 4 / 5 |
Total | 14 / 20 Passed |