Content
92%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.
A high-quality editorial skill body: dense with user-specific calibration facts, fully actionable specifications for every common request, an explicit calibration feedback loop with pre-draft validation, and clean one-level reference splitting. The only measurable weakness is minor redundancy between the curation and drafting sections.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Nearly every sentence carries non-derivable calibration data (rejected jokes, banned phrases, the 8.75/10 bar, confirmed vs hypothetical preferences) and nothing explains concepts Claude already knows. Only minor trims exist — e.g. the one-company-experience caution appears in both the curation and drafting sections — so it sits at efficient-with-minor-over-explanation rather than lean-every-token-earns-its-place. | 4 / 5 |
Actionability | For an instruction-only skill the guidance is fully concrete: exact per-candidate shortlist fields, exact feedback labels ("would share / useful to read / skip"), enumerated banned phrases ("caught my attention", "the difficult part is"), exact file paths and fields (content/profile.json, content/drafts.json with storyId and status), and exact queue states. Per the rubric's instruction-only note, absence of code is not penalized; the common cases are each fully specified. | 5 / 5 |
Workflow Clarity | Sections sequence research → shortlist → recommend → calibrate → develop → quality check, with an explicit feedback loop (offer candidates, collect feedback, update the calibration reference, test transfer on later batches) and explicit checkpoints: the pre-draft checklist ("Before presenting a draft, check whether..."), the deletion pass, and "Never mark an item published without evidence". No destructive or batch operation applies, so no cap; checkpoints are explicit rather than implicit, keeping it above the 4 anchor. | 5 / 5 |
Progressive Disclosure | Both bundle references exist and are exactly one level deep, each linked with a stated purpose: "Read [editorial-calibration.md]... for accepted examples and unresolved preferences before curating or drafting" and "The accepted direction is the example in [preferred-example.md]". Examples and calibration history are appropriately split out of a well-sectioned body with clear headers, matching the clear-overview/well-signaled-references anchor. | 5 / 5 |
Total | 19 / 20 Passed |