Content
75%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 well-constructed skill body: a clear five-step workflow with explicit defaults and fallbacks, concrete rendering specs, and properly signaled one-level-deep references that exist in the bundle. The main improvements are removing the duplication between Step 3's inline payoff formulas and strategies.md, and making the Step 4 JS template fully executable rather than skeleton stubs.
Suggestions
Deduplicate Step 3: keep only one inline worked example (e.g., butterfly) and point the rest to 'references/strategies.md', which already contains the identical formulas.
Make the Step 4 code template reference concrete implementations — e.g., 'include normCDF/bsPut/bsCall verbatim from references/bs_code.md' instead of stub bodies — so the widget script is copy-paste ready.
Add a lightweight verification step after Step 4 (e.g., confirm the widget rendered and sliders update the stats cards) to close the workflow's only checkpoint gap.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and efficient — field-extraction tables, terse formulas, explicit slider ranges and chart specs — with no tutoring on concepts Claude already knows. However, payoff formulas for butterfly, vertical, and iron condor are inlined in Step 3 while 'references/strategies.md' holds the same detail, a minor duplication that could be trimmed. | 4 / 5 |
Actionability | Mostly executable: exact SPX probe command, precise slider ranges (IV 5–80% step 0.5, DTE 0–90), chart color/line specs, and a full defaults table. The JS code template in Step 4 is skeleton pseudocode ('function normCDF(x) { /* Horner approximation */ }', 'function update() { ... }'), which is partially justified ('adapting ... per strategy') and backed by the copy-paste 'references/bs_code.md', so only a minor gap remains. | 4 / 5 |
Workflow Clarity | A clear 5-step sequence (extract → identify → compute → render → respond) with decision guidance for the riskiest step — spot price fallback ('Fall back to the middle strike only when no quote is available, and tell the user you did') and error handling in the probe. Missing only an explicit check that the rendered widget loaded correctly, so it falls just short of the validation-checkpoint anchor. | 4 / 5 |
Progressive Disclosure | Good structure: an overview body with a strategy-identification table pointing to 'references/strategies.md', a clearly labeled 'Reference Files' section listing both existing files with one-line descriptions, and references exactly one level deep (verified: both referenced files exist and contain the promised content). The inlining of several strategy payoff formulas that duplicate strategies.md keeps it below the cleanly-split top anchor. | 4 / 5 |
Total | 16 / 20 Passed |