Content
71%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 well-organized, lean command entry point whose diagnostic half (score/rescue) is genuinely executable thanks to concrete commands, sample JSON, and benchmark tables. Its weaknesses are externalized substance — the plan template and the scorer script are pointed at but not bundled or inlined — and prose-implied rather than explicit sequencing for the plan mode.
Suggestions
Ship the scorer script (or its essential logic) inside the skill's bundle, or inline the scoring formula/benchmark logic in SKILL.md so the score/rescue workflow works when the external marketing-skill repo is absent.
Add one sample scorer output block (per-stage rates vs. benchmark + named bottleneck) so Claude knows exactly how to interpret results and what the ranked fixes should look like.
Convert the `plan` mode's prose description into an explicit numbered sequence (intake → lock promise/format → backward sizing → runway → show-up/live-to-close/follow-up → deliverable via template), including the 'required visits > reachable audience' checkpoint as an explicit validation step.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and purposeful — intake as a 3-row table, benchmarks as a table, a worked funnel-math example, and runnable commands with no space wasted explaining what a webinar is. Minor trimmable padding exists: the 7-line 'Trigger Phrases' section largely restates the frontmatter description, and 'Anti-Patterns Rejected' slightly overlaps the funnel-math rationale. It sits above anchor 3 (which tolerates genuinely unnecessary explanation) but below anchor 5 because a few sections restate the same idea ('Treats a webinar as a funnel' appears in both the framing and the anti-patterns). | 4 / 5 |
Actionability | The scorer workflow gives copy-paste-ready commands ('python3 marketing-skill/skills/webinar-marketing/scripts/webinar_funnel_scorer.py funnel.json', stdin piping, no-args demo) plus a complete sample input JSON with required/optional fields, and the backward funnel math is a concrete worked example (20 SQOs → ~1,425 visits). It is not a 5 because the `plan` and `evergreen` modes stay abstract ('Walks the intake, locks the promise + format...') and defer entirely to a template file that is not in the bundle, and no sample scorer output is shown despite it being described as the diagnostic centerpiece. | 4 / 5 |
Workflow Clarity | Modes are enumerated with exact invocations, a 3-question intake defines the entry decision, and the workflow section sequences score/rescue correctly (run scorer → read per-stage rates vs. benchmark → named bottleneck → ranked fixes). The funnel math includes a real checkpoint ('If the required visits exceed the reachable audience, fix the goal, format, or promotion budget now'). Not a 5: the plan mode's multi-step flow (intake → promise/format → sizing → runway → deliverable) is implied in prose rather than explicitly sequenced, and there is no validate-and-retry loop around the scorer output (e.g., what to do when input JSON is malformed or a stage rate is missing). | 4 / 5 |
Progressive Disclosure | Section organization is clean (When To Run / When NOT To Run / Modes / Intake / Workflow / Benchmarks), but the skill has no bundle at all: the entire workflow depends on external files — the scorer script, the webinar plan template, and the sibling `webinar-marketing` SKILL.md — none of which exist under any referenced path in this bundle, and none of the core scorer logic or template structure is inlined as a fallback. That is 'structure present but the content that should back the references is absent,' fitting anchor 3 rather than anchor 4 (where references are mostly clear AND the supporting placement works). | 3 / 5 |
Total | 15 / 20 Passed |