Content
77%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.
An exceptionally actionable and well-gated migration workflow — every step has executable commands, explicit validation gates, and hard-won project-specific constraints that reviews repeatedly proved load-bearing. The weaknesses are token efficiency (duplicated explanations and anecdotal padding) and progressive disclosure (a monolithic body carrying reference material inline, with skill-local helper files referenced but absent from the bundle).
Suggestions
Deduplicate the check-vs-typecheck explanation in §6 — the ~690-file sync warning is made twice; state the rule and its reason once and reference it.
Move the historical/reference material (the 2026-08-21 database cutover, the attribution/backfill naming scheme, and the dated audit anecdotes) into a reference file linked from SKILL.md, keeping the body focused on the workflow rules those stories justify.
Trim each war-story passage to its actionable rule plus one concrete example (e.g., the Bulk Image Manager gap list can collapse to the pattern and a single instance), cutting the body's token cost without losing the constraint.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body assumes Claude's competence throughout — no basic concepts are explained — but it could be tightened: the check-vs-typecheck point is explained twice in §6 ("the sync half writes ~690 files into the directory the dev server watches" and later "which regenerates ~690 files under `.svelte-kit/`"), and dated war-story passages ("That collision froze a whole session's worth of work before it was diagnosed") pad rules that could stand alone. This fits the 3 anchor (mostly efficient, some unnecessary explanation); it is above 2 because nearly all content is non-obvious project-specific knowledge, and below 4 because the duplication and narrative padding are more than minor. | 3 / 5 |
Actionability | The guidance is fully executable and copy-paste ready: `node .claude/skills/retool-migration/extract.mjs "<export.json>" --queries`, the `getModeratorDb()` Kysely snippet, the `NAVIGATION` TypeScript block, `pnpm --filter ./apps/moderator run typecheck`, and concrete grep/sed recipes for endpoint discovery. Specific examples cover the common cases, matching the 5 anchor. | 5 / 5 |
Workflow Clarity | A clearly sequenced §0–§7 workflow with explicit validation gates for batch/destructive operations: the coverage gate ("Diff that against your classification. Anything unaccounted for is a gap, not a judgement call"), five named review passes, typecheck/check commands, and "Read the WARNING lines, not just ERROR". Feedback loops (fix findings, re-run reviews, re-validate) are present throughout, matching the 5 anchor. | 5 / 5 |
Progressive Disclosure | External references are clearly signaled and one level deep (CLAUDE.md, retool-db-cutover.md, report-queue-pattern.md), but the SKILL.md itself is a ~435-line monolith with reference-type material inlined — cutover history ("Since 2026-08-21 it points at the moderator database"), attribution/backfill naming policy, and dated audit anecdotes — that belongs in reference files. Additionally, the helper files the instructions depend on (`extract.mjs`, `retool-db.mjs`, `CLICKUP-SCOPE.md`, `MIGRATIONS.md`) are not present in the evaluated bundle (no references/, scripts/, or assets/ directories exist). This fits the 3 anchor (some structure, content that should be separate is inline); it is above 2 because sections are well-organized and links are clearly signaled, and below 4 because the missing bundle files and inlined reference material are more than minor organization gaps. | 3 / 5 |
Total | 16 / 20 Passed |