Content
86%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 tight, well-structured instruction skill with concrete commands and a sensible security guardrail around community skills. Its one real defect is the truncated final workflow step, which cuts off mid-sentence and leaves the name-conflict guidance incomplete; the trust-review step also lacks specifics on what to inspect.
Suggestions
Complete the truncated step 5 — the sentence ends at "take precedence over global `~/.codewhale/skills`, with" and never states the resolution.
Specify what to look for when reviewing a community SKILL.md before trusting it (e.g., prompt-injection in instructions, destructive or network-capable scripts, unexpected file writes).
Add a recovery step for failed installs (e.g., re-run /skill install, check the registry name, fall back to manual clone into ~/.codewhale/skills).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is ~30 lean lines: a command list, a 5-step workflow, and a one-line security guardrail. It assumes Claude's competence — no explanation of what skills are, what GitHub is, or how installation works — so every token earns its place. Not below 5 because there is no padding or over-explanation anywhere (the opening sentence slightly duplicates the description but is a standard, minimal pattern). | 5 / 5 |
Actionability | Concrete, executable commands are given ("/skill install github:<owner>/<repo>", "/skill trust <name>") and the workflow maps to them. Not 5: step 5 is truncated mid-sentence ("take precedence over global `~/.codewhale/skills`, with"), leaving an unusable instruction, and step 3 says to "inspect the SKILL.md" without stating what to look for (prompt injection, destructive scripts, network calls). Not 3: the guidance is genuinely executable apart from that one broken step. | 4 / 5 |
Workflow Clarity | The 5-step workflow has a clear sequence and some checkpoints: review-before-trust in step 3 and a verify-installed check in step 4 ("restart the session if the skill does not appear"). Not 5: the truncated step 5 makes the name-conflict resolution incoherent, and there is no explicit feedback loop for the install failure case (e.g., what to do when /skill install errors). Not 3: sequence and several checkpoints are clearly present, not merely implied. | 4 / 5 |
Progressive Disclosure | The skill is under 50 lines with no need for external references — no references/, scripts/, or assets/ bundle exists, and nothing in the body points to files that should exist. Two well-labeled sections (## Commands, ## Workflow) plus a scoped closing guardrail make navigation trivial, matching the simple-skill exception for progressive disclosure. | 5 / 5 |
Total | 18 / 20 Passed |