Content
46%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 delivers many concrete, executable commands and benefits from numbered sections and a Quick Reference table, but it is a monolithic 900-line catalog with duplicated tips, provenance/date metadata, and no progressive disclosure or reference files. As a context-window consumer it reads as a reference dump rather than a lean, navigable skill.
Suggestions
Split the 60 sections into one-level-deep reference files by theme (e.g., parallelism/worktrees, hooks & permissions, scheduled tasks, voice/mobile) and keep SKILL.md as a concise overview with clearly signaled links, replacing 'All tips are contained in this file'.
Merge the duplicated tips (/btw 33/54, /batch 30/56, /loop 31/48, verification 14/51) and move the 'Parts' thread provenance, per-tip source URLs, and date/version stamps into a separate SOURCES.md so they don't consume context on every load.
Add a short 'How to use this skill' section giving a sequence for matching a user's question to the relevant tips (e.g., identify topic -> check Quick Reference -> present the tip with its command), which would also give the update-check procedure a defined place in the flow.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~900-line body includes clear padding: a 10-line provenance 'Parts' listing with per-thread dates, per-tip source URLs, and four near-duplicate tip pairs (/btw in sections 33 and 54, /batch in 30 and 56, /loop in 31 and 48, verification in 14 and 51), plus time-sensitive dates and versions (v2.1.74+, Jan 2 2026) with no 'old patterns' framing. Not 1 because individual tips are tight and free of basic-concept explanation; not 3 because the duplicated sections, provenance metadata, and date-stamped content are pervasive rather than occasional. | 2 / 5 |
Actionability | Most tips give concrete, executable guidance: `git worktree add .claude/worktrees/my-worktree origin/main`, `claude --worktree my_worktree --tmux`, full PostToolUse hook JSON, `claude -p "summarize this codebase" --output-format=stream-json --verbose --bare`. Not 5 because a meaningful minority of tips are advisory with no concrete form ("Write detailed specs", "Challenge Claude", "Use Claude for Learning"), leaving gaps in executable coverage. | 4 / 5 |
Workflow Clarity | The content is a 60-section topic catalog rather than a sequenced workflow; there is no instruction on how to select or apply tips to a user's situation, and the only internal procedure (the UPDATE CHECK version-compare block) is a single unvalidated fetch-and-compare step. Not 2 because individual tips do present coherent sequences (plan mode -> refine -> auto-accept -> 1-shot) and the Quick Reference table maps topics to actions; not 4 because no overarching workflow or checkpoints structure the skill's own operation. | 3 / 5 |
Progressive Disclosure | Everything is inlined in one ~900-line file with zero reference files ("All tips are contained in this file — do not fetch from the website"), far past the 50-line threshold where inline-only structure could score 5. Not 1 because the 60 numbered sections and Quick Reference table provide real navigation structure; not 3 because there are no references at all and content that clearly belongs in separate files (e.g., topic-grouped tip collections, the Parts provenance) is fully inlined. | 2 / 5 |
Total | 11 / 20 Passed |