Content
73%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-sequenced orchestrator skill with outstanding workflow clarity — explicit checkpoints, resume logic, and review feedback loops — and largely concrete, executable sub-skill invocations. Its weaknesses are moderate padding (patent-law fact tables and three overlapping pipeline diagrams) and a vague, jargon-heavy reviewer-routing preamble that lacks concrete commands, plus a dangling reference to a shared file absent from the bundle.
Suggestions
Move the invention-vs-utility-model specification tables into a reference file (e.g., references/patent-types.md) and keep only the decision-relevant constraints (CN-only, apparatus claims only for utility models) inline, cutting the restated patent-law facts Claude already knows.
Replace the three overlapping pipeline diagrams (linear chain, branch tree, and 'Composing with Other Workflows') with a single compact diagram, or fold the entry-points list into a few bullet lines.
Make the reviewer-routing preamble concrete: give the actual reviewer_route.py invocation and an example .aris/reviewer.json, or move the routing details into the referenced reviewer-backends.md file (which should be included in the bundle so the link resolves).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly operational, but the patent-type tables restate well-known patent law facts Claude already knows ("20 years from filing date", "formal examination only", "6-8 months to grant"), and the pipeline structure is diagrammed three times (the linear chain, the branch tree, and again in "Composing with Other Workflows"). This fits anchor 3 (mostly efficient but includes some unnecessary explanation or could be tightened) rather than anchor 2, since no section is purely concept-teaching and most guidance is skill-specific. | 3 / 5 |
Actionability | Phases 0-5 provide copy-paste-ready sub-skill invocations (e.g., `/prior-art-search "patent/INVENTION_BRIEF.md"`), a complete PATENT_STATE.json schema, exact checkpoint message templates, and concrete override syntax. It falls short of anchor 5 because the reviewer-routing preamble ("resolve invocation → `.aris/reviewer.json` → Codex default ... with `reviewer_route.py`") gives no executable command or example, leaving a gap in a core review phase — anchor 4 (mostly executable with minor gaps). | 4 / 5 |
Workflow Clarity | The workflow is a clearly sequenced six-phase pipeline with explicit STOP-and-wait checkpoints, enumerated user options ("go" / adjustments / "stop"), state persistence after every phase, 24-hour resume logic, and a two-round examiner review-revision feedback loop. This matches anchor 5 (clear sequence, explicit validation, feedback loops); anchor 4 would require missing or implicit checkpoints, which is not the case. | 5 / 5 |
Progressive Disclosure | The document is well-sectioned with clear headers and its one detailed reference ("../shared-references/reviewer-backends.md") is clearly signaled at the top and one level deep. It stops short of anchor 5 because the referenced shared file is not present in the bundle, the bundle contains no references/ files at all, and inline content such as the patent-type specification tables and the deliverables table could appropriately live in separate reference files. | 4 / 5 |
Total | 16 / 20 Passed |