skill-social-performance-review · Monthly Social Review
Monthly social performance review: generates a client-readable monthly review — account snapshot, best/worst post analysis, content-pillar and format breakdowns, key insights, and next-month action items ranked by expected impact.
Computation is script-deterministic (per-post scoring, top/bottom lists, month-over-month, internal 1–10 score); the LLM only interprets. Adapts to three data qualities (CSV export / screenshots / verbal), never blocked by missing data — limitations are flagged. Paid-promotion posts are excluded from organic baselines; next-month advice must be concrete and executable.
Example invocation: "Generate this month's social performance review."
Full brief
Positioning
skill-social-performance-review is a cross-platform portfolio-level monthly review: analyzing last month's social performance and producing a client-readable report with next-month action items. Not publish-log four-dimension attribution (that's skill-publish-analytics).
Core capabilities
- Data-quality tiers: full (CSV export) / partial (screenshots or top/bottom lists) / minimal (verbal) — analysis never stalls for missing data; reports state data source and quality tier up front.
- Script-deterministic computation: standardized data becomes JSON, handed to
scripts/review.pyfor per-post engagement rates and composite scores, top3/bottom3, pillar and format aggregation, month-over-month, weighted internal score; the LLM interprets, never hand-computes. - Seven analyses: account snapshot / best posts / worst posts / pillar performance / format performance / opening analysis (reading the copy's first line) / posting rhythm.
- Key insights + next-month advice: 2–4 causality-connected insights; 3–5 impact-ranked recommendations, each with "what to do / data evidence / how to execute" — concrete and executable (e.g. "carousels from 2 to 4 per month, focused on the best-performing pillar").
- Review compounding: this month's Top 3 appended to best-performers.md, monthly summary logged to review-history.md for month-over-month comparison.
- Competitor watch (optional): 4–6 comparison points when competitor accounts exist and Playwright/Firecrawl MCP is configured; skipped and noted otherwise.
Workflow
- Prep (brand style, last month's schedule, historical top performers, last month's review)
- Gather (month, platforms, data source, business background; export guidance when data is missing)
- Standardize (CSV/screenshots/verbal unified into metrics; paid-promotion posts excluded from organic baseline)
- Script computation + seven analyses (interpret script results)
- Insights & advice (key insights, next-month actions, schedule adjustments)
- Output report (per template) + update compounding context + handoff notes
Inputs & outputs
| Input | Notes |
|---|---|
| Review month | Required |
| Platforms | Xiaohongshu/Douyin/Bilibili/Weibo/WeChat OA (multi-select) |
| Data source | CSV export / screenshots / verbal (descending priority) |
| Business background | Optional — special events, promotions, paid boosts |
Output: structured monthly review report (overview, best/worst posts, pillar and format breakdowns, insights, next-month advice), saved to outputs/复盘主题/.
Boundaries with adjacent skills
| Skill | Lane |
|---|---|
| skill-social-performance-review (this) | Cross-platform portfolio-level monthly review + next-month actions |
| skill-publish-analytics | Four-dimension attribution from the publish log |
| skill-content-postmortem | Single-post win/loss review / hit-formula extraction |
Fit
- Monthly operations summary and next-month planning
- Client-readable monthly report deliverables
- Finding what works and what fails at the portfolio level
Before you start
- Prepare CSV exports (platform creator backends) or screenshots; verbal top/bottom lists work as a minimum.
- Confirm which posts were paid promotions (they pollute the organic baseline — must be excluded).
skill-social-performance-review is part of the Aiglade Skill library. Invoke it from the Aiglade chat box in plain language.