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.
The body is a tight, actionable rulebook: 35 concrete checkable rules with severities, a complete JSON output contract, and a clear quality gate, with no bundle files to navigate. Its main weaknesses are the unspecified computation for the numeric score and anti_slop_axes ratings and a merely implicit execution sequence.
Suggestions
Define how the numeric 'score' is computed (e.g., start at 100, −10 per Errore, −5 per Avviso) and how rule violations map to each anti_slop_axes rating, so the output contract is fully executable.
Add a brief 3–4 step execution sequence (read README → apply rules by severity → compute score/axes → emit JSON report) to make the workflow explicit rather than implied.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and rule-focused with almost no concept explanation Claude already knows — each of the 35 rules is a terse table row and the banned-words list is execution data, not padding. The only trimmable redundancy is the 7-axis table partially restating rules 11–20, which keeps it just below the lean 5 anchor. | 4 / 5 |
Actionability | Rules are concretely checkable (e.g., 'Nessun riferimento a `0.0.0.0`', an explicit banned-word list, severity levels, and a fully specified JSON output example), making guidance mostly executable. Minor gaps remain: the numeric 'score' (e.g., 85) and the anti_slop_axes ratings have no stated computation rule, so it falls short of fully copy-paste-ready 5. | 4 / 5 |
Workflow Clarity | As a simple single-pass lint the action is unambiguous in shape (apply the rule tables, emit the JSON report), and the read-only nature avoids the destructive/batch validation cap. However, the execution sequence (read → check by severity → score → report) is only implicit and score computation is unspecified, placing it at 4 rather than 5. | 4 / 5 |
Progressive Disclosure | The single file (~100 lines) is well organized into clearly signaled sections (fgt-sdk rules, three anti-slop rule sets, output, quality gate, triggers) with all content needed at execution time, so inlining is appropriate. It exceeds the under-50-line simple-skill bar and the rule tables could arguably split into reference files, so it sits at 4 rather than 5. | 4 / 5 |
Total | 16 / 20 Passed |