skill-post-scorer · Draft Performance Scoring
A pre-publish quality skill that scores a social post draft's engagement potential across five dimensions and outputs a structured scorecard. It judges the finished draft (pre-publish) — not untried topic ideas (that's skill-topic-evaluator) and not monthly post-hoc reviews.
Scoring is grounded in historical data: when historical post data exists, a script first computes the account's top-performing posts and their patterns as the scoring baseline; without historical data it falls back to generic benchmarks, clearly labeled as such.
Example invocation: "Score this draft for me," then paste the post copy.
Full brief
Positioning
skill-post-scorer is a pre-publish gatekeeper: it scores a social post draft's engagement potential and outputs a structured five-dimension scorecard with a verdict. It answers "is this draft worth publishing, and what should change" — draft quality, not topic potential or attribution review.
Core capabilities
- Five-dimension scoring: hook strength, voice match, value density, structure & format, publish readiness — 1–10 each, 50 total.
- History-grounded: with historical post data, a script computes top posts by engagement score (likes + comments×3) and their patterns as the scoring baseline; the LLM only extracts the pattern features.
- Graceful fallback: without historical data it uses generic benchmarks, explicitly labeled "based on generic benchmarks".
- Verdict and rewrite: four tiers by total score (40+ publish as-is / 30–39 publish after fixes / 20–29 needs revision / <20 rewrite), then an offered targeted rewrite of the weakest dimension.
- Scoring discipline: honest scores, no flattery; nothing scores 8+ unless it genuinely matches the top-10% pattern.
Workflow
- Capture the draft: read pasted copy or a file in
outputs/; target platform optional - Load scoring data: profile history → user-provided data → generic benchmarks, in priority order
- Analyze top performers: the script identifies top posts; the LLM extracts shared hook types, pacing, format traits, CTA types
- Score across five dimensions per the rubric
- Output the scorecard: code-block format with per-dimension bars, total, verdict, and weakest-dimension fixes
- Offer rewrite: with consent, rewrite the weakest part and re-score for comparison
Inputs & outputs
| Input | Notes |
|---|---|
| Draft copy | Required — paste the text |
| Draft file path | Or point to a file in outputs/ |
| Target platform | Optional — Xiaohongshu, Douyin, Weibo, Zhihu, WeChat, Bilibili, etc. |
| Historical post data | Optional — CSV/Excel/JSON exports; sharper scoring with real data |
Output: a code-block scorecard (five scores + total + verdict + improvement notes).
Boundaries with adjacent skills
- skill-topic-evaluator: scores untried topic ideas; this skill scores finished draft quality.
- skill-social-performance-review: monthly post-publish portfolio review; this skill is pre-publish, single-draft.
Fit
- Quick pre-publish check: is this draft worth publishing, where's the bottleneck
- Comparing several versions of one topic and picking the highest score
- Benchmarking against your own account's top patterns once history accumulates
Before you start
- Paste the draft and go; unspecified platforms fall back to generic social best practices.
- Historical post exports (CSV/Excel from creator dashboards) make scoring data-driven; without them the scorecard states the generic baseline.
- No API key or fees involved.
skill-post-scorer is part of the Aiglade Skill library. Invoke it from the Aiglade chat box in plain language.