Content
67%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 dense, contract-style skill body with genuinely actionable parameters (exact flags, numeric retry/circuit-breaker policy, templates), a clear phased workflow with explicit verification gates, and well-signaled one-level-deep references. Its main weaknesses are repetition — the 5-axis check and pair-mode contract restated multiple times, plus a 48-line duplicative capabilities comment — and the absence of any executable code example or verifiable bundle files behind the many reference/ pointers.
Suggestions
State the 5-axis Impact Scope Check once (e.g., in the Workflow VERIFY row with the YAML template) and reference it by name from Core Contract, Always, and Output Requirements instead of restating the full definition in each place.
Remove or drastically shrink the CAPABILITIES_SUMMARY HTML comment block, which duplicates body content line-for-line and costs ~48 lines of token budget on every skill load.
Add one explicit failure-recovery loop to the Workflow (e.g., "if VERIFY finds an axis NEEDS-REVIEW or a gate fails: fix, re-run the verification, only then advance to PRESENT") to reach the top workflow-clarity anchor.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body almost never explains concepts Claude already knows — it even states "General language syntax and design-pattern tutorials are intentionally not stored in this skill" — but it is noticeably padded by repetition: the 5-axis Impact Scope Check is fully restated in at least five places (Core Contract bullet, Always list, VERIFY workflow row, Output Requirements bullet, and a dedicated YAML template section), and the ~48-line CAPABILITIES_SUMMARY HTML comment block duplicates body content nearly line-for-line. This fits the 3 anchor's "mostly efficient but... could be tightened" better than the 4 anchor's "minor instances". | 3 / 5 |
Actionability | Concrete, executable specifics are dense throughout: exact compiler flags ("noUncheckedIndexedAccess, exactOptionalPropertyTypes, noPropertyAccessFromIndexSignature"), numeric resilience policy ("429 backoff with Retry-After, 5xx exponential 3-5 attempts", "open after 5 failures in 60s (payment <= 3, search <= 10)"), a copy-paste ImpactScopeReport YAML template, an exact journal row format, and a dispatch allowlist. It stops short of the 5 anchor because there is no executable code example anywhere and many specifics defer to reference/ files that are not part of this bundle. | 4 / 5 |
Workflow Clarity | SURVEY → PLAN → BUILD → VERIFY → PRESENT is laid out in a phase table with per-phase focus, key actions, and read-triggers, and VERIFY carries explicit validation checkpoints — "never declare 'done' without all 5 axes verified or explicitly N/A", the CODE_QUALITY_GATE, "SEC: risk blocks completion", and per-recipe acceptance gates. The gap versus the 5 anchor is that no explicit error-recovery loop states what happens when a gate or verification fails (the closest is "Fix root causes; never suppress symptoms" and the NEEDS-REVIEW → ripple escalation). | 4 / 5 |
Progressive Disclosure | References are one level deep and well signaled — a full index ("reference/reference-index.md"), a recipes index, per-row "Read this when" triggers, "load only its 'Read First' files", and "Read only the files required for the current decision" — which is close to the 5 anchor. It lands at 4 because no bundle files ship alongside SKILL.md (the referenced reference/ and _common/ paths cannot be verified to exist), and the bulky inline CAPABILITIES_SUMMARY comment plus the ~25-bullet Core Contract keep overview-level content in the main file that could live in reference files. | 4 / 5 |
Total | 15 / 20 Passed |