Content
88%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 an exemplar of actionable, project-specific skill writing: concrete APIs, naming registries, privacy rules grounded in real incidents, and a numbered workflow with live verification and feedback loops. Its only notable weaknesses are mild repetition across the One-idea/Rules/Workflow/Anti-patterns sections and a SKILL.md that carries more inlined doctrine (especially session replay) than a lean overview-plus-references split would.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and project-earned — no explanation of what PostHog is or how analytics works — with nearly every token carrying non-obvious, hard-won guidance. Not anchor 5 because there is mild repetition: the "One idea" paragraph restates Rule 1, and the Workflow and Anti-patterns sections restate the six rules (efficiently, via "(Rule N)" cross-references, but still overlapping). | 4 / 5 |
Actionability | For an instruction-only skill the guidance is fully concrete: exact hook (`usePostHogClientCapture()` → `capture(eventName, props?)`), exact registry mechanism and typecheck enforcement, exact naming conventions with real examples (`saved_views:view_selected`), exact prop rules (`objectAdjective`, `is`/`has`, `Date` suffix), and an exact verification method (Playwright spy on `window.posthog.capture` or the ingest POST, with three named assertions). Per the rubric's code-vs-instruction note, absence of code is not penalized when guidance is this actionable. | 5 / 5 |
Workflow Clarity | The Workflow section gives a clear 5-step sequence (plan → register → wire → verify live → PR) with an explicit validation step carrying three concrete assertions (fire exactly once, dimension correct, no raw values), a feedback loop ("Fix any pre-existing leaks you pass"), a checklist section of anti-patterns, and a separate branch for replay-only changes. This matches anchor 5's sequence-plus-validation-plus-feedback shape. | 5 / 5 |
Progressive Disclosure | Both referenced bundle files exist, are one level deep, and are clearly signaled with read-when guidance ("Read when designing a new taxonomy or debating naming/scope", "Read when hardening the pattern or adding global dimensions"), matching the top of the anchor range. Held at 4 rather than 5 because the body itself inlines the full six-rule doctrine and an entire session-replay privacy sub-skill (~80 lines) that could partly live in a reference, making SKILL.md heavier than a clear overview. | 4 / 5 |
Total | 18 / 20 Passed |