Content
78%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 lean, well-structured atom with concrete formulas, explicit inputs, and a defined output contract. Its main weaknesses are the missing validation checkpoints in the workflow and a Load pointer to a reference file that is not part of this bundle, which leaves one step's detail unresolvable.
Suggestions
Ship the revenue-join.md reference inside this skill's own references/ directory (or inline the like-for-like calculation) so the 'as in the reference' dependency resolves when the atom is used standalone.
Add a light validation step, e.g. sanity-check that store/CRM weeks align with spend weeks and that claimed totals exceed store totals before computing the overclaim ratio.
Clarify where 'the story the marketer picked' is read from (which section of last week's report) so step 5's carry-forward rule is fully executable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~20-line body is pure instruction — Load, Inputs, six numbered Do steps, Output — with zero padding and no explanation of concepts Claude already knows. Every line tells the agent something it could not infer, matching 'lean and efficient; every token earns its place'. | 5 / 5 |
Actionability | Concrete guidance throughout: exact formulas ("claimed purchases ÷ store orders", "Δ new-customer revenue ÷ Δ spend"), named metrics with comparison baselines (break-even/allowable/target, last week, 4-week average), and a specified output location. Not 5 because key detail ("like-for-like as in the reference") is deferred to ../analyze-performance/references/revenue-join.md, which is not present in this bundle, leaving that step unresolvable standalone. | 4 / 5 |
Workflow Clarity | Steps 1-6 give a clear, well-ordered sequence (truth table → core metrics → overclaim → marginal → conflicting-story handling → B2B variant) with explicit Inputs and Output sections. Not 5 because there are no validation or sanity-check checkpoints (e.g., confirming the join keys or that store numbers are complete before computing ratios); not 3 since the sequence is complete with no real gaps, and no destructive/batch cap applies. | 4 / 5 |
Progressive Disclosure | Sections (Load/Inputs/Do/Output) are well organized and the single reference is clearly signaled, but the referenced file ../analyze-performance/references/revenue-join.md does not exist in this bundle (no references/, scripts/, or assets/ directories), so the disclosure chain is broken — the deferred detail is unreachable. This sits between 'references present but not clearly signaled' (3) and 'minor organization gaps' (4); the dangling pointer is more than minor. | 3 / 5 |
Total | 16 / 20 Passed |