Content
85%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 content is exceptionally actionable — executable commands, decision trees, checklists, validation, and rollback for every mode — with only minor token waste. Its structural weakness is progressive disclosure: the body repeatedly points to references/karpenter.md, which is absent from the bundle, while ~490 lines of per-mode detail remain inlined in SKILL.md instead of being split out.
Suggestions
Create the referenced references/karpenter.md (NodePool design, EC2NodeClass, IAM, Troubleshooting, Disruption strategy, Cost strategy, Migration, Upgrade sections) or remove the six dangling 'Reference:' pointers — as delivered, every pointer leads to a nonexistent file.
Move the stable reference material — the v0.x→v1.x field-mapping tables, the audit-mode CloudTrail/jq queries, and the IAM policy details — into references/karpenter.md so SKILL.md stays a navigable overview of the seven modes.
Trim duplicated framing: the opening sentence 'Design, install, debug, review, plan capacity, audit, migrate, and upgrade Karpenter on EKS' repeats the frontmatter description and can be cut.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and almost entirely operational (commands, decision trees, checklists), with little explanation of concepts Claude already knows. Minor trimmable padding exists: the opening line "Design, install, debug, review, plan capacity, audit, migrate, and upgrade Karpenter on EKS" restates the description verbatim, and asides like "useful for cost estimates, pre-rollout capacity checks, and catching NodePool mismatches before they page you" are motivational rather than instructional. Fits 'Efficient; minor instances of over-explanation that could be trimmed' rather than the score-5 anchor's 'every token earns its place'. | 4 / 5 |
Actionability | Fully executable, copy-paste-ready guidance throughout: "crane ls public.ecr.aws/karpenter/karpenter | grep -E '^[0-9]+\.[0-9]+\.[0-9]+$' | sort -V | tail -5", "kubectl apply --dry-run=server -f ec2nodeclass.yaml", complete CloudTrail/jq pipelines, and concrete v0.x→v1.x field-mapping tables. Placeholders like <cluster-name> are appropriate parameterization, not pseudocode; matches the anchor "Fully executable; copy-paste ready code or commands; specific examples cover the common cases". | 5 / 5 |
Workflow Clarity | Each mode has an ordered, evidence-first sequence with explicit validation and feedback loops: debug insists "Collect evidence first — do not suggest fixes before seeing output"; migrate has a pre-flight checklist ("all must pass before proceeding") and warns "If PDBs block drain, the command will wait — do not force"; upgrade includes a post-upgrade validation block and rollback; generate ends with dry-run validation. Destructive/batch operations all carry validation plus Rollback sections — matches the anchor "Clear sequence with explicit validation steps; feedback loops for error recovery; checklists for complex processes". | 5 / 5 |
Progressive Disclosure | Every mode ends with a clearly signaled one-level pointer ("Reference: references/karpenter.md → NodePool design, EC2NodeClass, IAM"), but no references/ directory (or any bundle file) exists in the skill — all six pointers dangle. Additionally, ~490 lines covering seven modes with large inline detail (field-mapping tables, CloudTrail queries, full decision trees) is content that belongs in the referenced file. This sits between the score-3 anchor ("content that should be separate is inline", references unreliable) and score-4 ("minor organization gaps") — a systematically missing reference target plus heavy inlining is more than a minor gap, so 3. | 3 / 5 |
Total | 17 / 20 Passed |