Content
71%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 highly actionable pattern catalog with excellent copy-paste-ready templates and embedded validation for the install and multicluster workflows. Its weaknesses are padding (generic Instructions / Do-not-use sections, an ASCII diagram of known concepts, a pinned version string) and a fully monolithic single-file structure with no progressive disclosure despite being well over 50 lines.
Suggestions
Delete the content-free 'Do not use this skill when' and 'Instructions' sections, which contain no Linkerd-specific information, and remove or shrink the ASCII architecture diagram of concepts Claude already knows.
Split the longer deep-dive material (AuthorizationPolicy block, multi-cluster setup) into files under references/ and keep SKILL.md as a concise overview with clearly signaled one-level-deep links.
Add inline post-apply verification to the YAML templates (e.g., 'linkerd viz routes deploy/my-app' after a traffic split, 'linkerd check --proxy' after injection) so each template has a validation checkpoint, and move the pinned version/SHA-256 into a clearly dated or version-agnostic form.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The templates and command sections are lean and code-dense, but the document carries pure padding: the "Do not use this skill when" bullets ("The task is unrelated to linkerd patterns") and the generic "Instructions" section ("Clarify goals, constraints, and required inputs") say nothing domain-specific, the ASCII architecture diagram re-explains concepts Claude already knows, and a pinned version "edge-26.11.1" plus full inline SHA-256 is time-sensitive content not placed in a deprecated section. It is not anchor 2 because the padding is a small fraction of the ~355 lines and the core content is highly efficient. | 3 / 5 |
Actionability | Every section provides copy-paste-ready material: a complete checksum-verified install script, seven concrete YAML/bash templates (injection, ServiceProfile with retry budgets, canary HTTPRoute, AuthorizationPolicy, multi-cluster), plus monitoring and debugging command lists. Specific examples cover the common cases end to end, matching the anchor-5 example. | 5 / 5 |
Workflow Clarity | The major workflows embed explicit checkpoints — "linkerd check --pre", "sha256sum --strict -c -", "linkerd check" after install, and "linkerd multicluster check" / "gateways" for verification — with a clear implicit sequence within each template. It is not anchor 5 because the YAML config templates lack inline post-apply validation (e.g., verify a traffic split with `linkerd viz routes`) and there are no error-recovery feedback loops; it is above anchor 3 because key sequences do include explicit validation steps rather than leaving them implicit. | 4 / 5 |
Progressive Disclosure | No bundle files exist and no internal references are used — all ~355 lines live inline in a single file, including self-contained deep dives (the five-resource AuthorizationPolicy block, multi-cluster setup) that could be split into reference files with the SKILL.md as an overview. Section headers are clear and navigation is adequate, which keeps it above anchor 2's "no section headers" case, but the total inlining of content that belongs in separate files matches anchor 3. | 3 / 5 |
Total | 15 / 20 Passed |