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.
An exceptionally lean, well-organized skill body with real guardrails and a sensible classify-to-QA workflow. Its two real defects are the missing `templates/` bundle directory the core flow depends on, and the absence of an explicit lint/QA failure feedback loop and any inline example email structure.
Suggestions
Create the `templates/` directory with one template per email class (invite, reminder, day-of, post-attend, post-no-show, nurture), or inline minimal skeletons — step 2 of the flow currently points at files that do not exist.
Add an explicit feedback loop after the lint/QA steps, e.g. "If lint or QA fails: revise, then re-lint before delivering."
Include one worked example inline (subject + preview text + short body) so the skill is self-sufficient even if templates are not loaded.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is ~55 lines of lean bullet points, a compact inputs table, and a numbered flow — it assumes Claude's competence, teaches nothing the model already knows, and every line carries non-obvious information (guardrails, thresholds like "subject ≤50 chars", delivery policy). Matches 'lean and efficient; every token earns its place'. | 5 / 5 |
Actionability | The flow is concrete (named email classes, a lint step against references/anti-slop.md, a QA checklist), but step 2 — "Pick template from `templates/` matching the class" — is not executable: no templates/ directory exists in the bundle, and the body includes no example email skeleton or subject-line pattern of its own. Missing key material for the core task places this at 'some concrete guidance but incomplete', below the mostly-executable bar of 4. | 3 / 5 |
Workflow Clarity | A clear six-step sequence (Classify → Pick → Fill → Lint → Deliver → QA) with guardrails for the risky operation ("do not send unasked", deliver as HubSpot draft). Validation checkpoints exist (lint step, QA step), but there is no explicit feedback loop (what to do when the lint or QA fails) — 'clear sequence with most checkpoints present; minor validation gaps'. | 4 / 5 |
Progressive Disclosure | Well-sectioned short body with a one-level-deep, clearly signaled reference (references/anti-slop.md, which exists and is substantive). However the other referenced path, `templates/`, does not exist in the bundle — a broken reference that keeps this from the clean 5 ('clear overview with well-signaled one-level-deep references') but still sits above 'references present but not clearly signaled'. | 4 / 5 |
Total | 16 / 20 Passed |