菜单
Use when scoring supplied conversation data for sentiment, themes and share of voice. Produces sentiment analysis with method, confidence and implications; use `meta-social-listening` when that neighbouring contract is the closer match.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/peterbamuhigire/social-media-skills --skill meta-sentiment-analysis命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
基于 SOC 职业分类
| name | meta-sentiment-analysis |
| description | Use when scoring supplied conversation data for sentiment, themes and share of voice. Produces sentiment analysis with method, confidence and implications; use `meta-social-listening` when that neighbouring contract is the closer match. |
| metadata | {"portable":true,"compatible_with":["claude-code","codex"]} |
Scope distinction: This skill covers the analytical methodology — how to score, calculate, and interpret sentiment data, and how to translate findings into strategic decisions. Use
playbook-sentiment-listeningfor the operational setup that generates the raw data. The two skills are complementary: listening provides the data; this skill provides the analysis.
meta-social-listening is not the closer route before proceeding.meta-social-listening when its narrower output is requested.| Artefact | Source/provider | Required? | If absent |
|---|---|---|---|
| Dated listening dataset, query taxonomy and sampling method | Client, approved systems, or dated platform exports | Yes | Stop the affected decision; request it or mark the field unknown and narrow the output. |
| Purpose, audience and approval boundary | Client brief or accountable owner | Yes | Return discovery questions; do not infer approval. |
| Artefact | Consumer | Acceptance condition |
|---|---|---|
| Sentiment analysis with method, confidence and implications | Client lead and next workflow owner | Every recommendation traces to an input, names an owner or next action, and marks assumptions and unassessed checks. |
| Evidence | Format | Acceptance condition |
|---|---|---|
| Decision and source register | Table in the deliverable | Each material claim records its source/date or is labelled unverified; missing evidence never becomes a pass. |
Read and search access to the supplied artefacts are required; calculation or file-rendering capability is optional. This is read-only by default: inspect and report without changing source records, accounts, skills or campaigns. Editing the deliverable requires explicit authorisation; publishing, production mutation, destructive action, spend, and certification claims require separate explicit authority and evidence.
If files, platform access, network, rendering, fonts, or calculation tools are unavailable, return the narrowest useful qualified sentiment analysis with method, confidence and implications. Mark each blocked check not assessed, state the consequence, and provide the exact evidence needed to resume. Never convert an unavailable check into a pass.
| Choice | Action | Failure or risk avoided |
|---|---|---|
| Dated listening dataset, query taxonomy and sampling method is current and attributable | Produce the full sentiment analysis with method, confidence and implications and cite the evidence used. | Decisions based on stale or unrelated evidence. |
| A material input is missing or contradictory | Stop that decision, request clarification, or issue a labelled partial result. | Fabricated precision and false confidence. |
The requested outcome belongs to meta-social-listening | Route there and hand over the verified inputs already collected. | Neighbour collision and duplicated work. |
meta-social-listening if its contract is closer.The output is client-specific, uses British English and the stated market/currency, distinguishes observed fact from inference, exposes gaps, and gives a checkable acceptance condition. Recommendations must be feasible within the confirmed budget, capacity and permissions.
not assessed, its risk and the recovery input.meta-social-listening into this workflow. Fix: route the neighbouring output and hand over verified inputs.Given verified dated listening dataset, the skill produces a sentiment analysis with method, confidence and implications with source dates and named assumptions. If that evidence cannot be accessed, it returns only the supported sections plus a recovery list; it does not fill gaps with East African defaults.
meta-social-listening for the neighbouring contract.anti-ai-slop during production.ai-slop-audit at the release checkpoint.Collect the following before generating any deliverable:
Choose the scoring method based on the client's tool budget and audience language mix. The hybrid approach is recommended for most East African clients.
Manual scoring is appropriate when the client has no budget for sentiment tools, when the audience writes primarily in Luganda or Swahili, or when monthly mention volume is below 500 across all platforms.
Process:
Four-category classification scheme:
Five negative sub-categories:
Time required: 30–45 minutes per platform per month. Budget this time into the monthly reporting cycle.
Automated scoring is appropriate when English-language mention volume exceeds 500 per month per platform, or when real-time monitoring is required.
Recommended tools by budget tier:
| Budget | Tool | Notes |
|---|---|---|
| Free | MonkeyLearn (basic API) | 300 queries/month free; sufficient for small accounts |
| USD 0–30/month | Mention (Starter) | English NLP; basic dashboard |
| USD 30–100/month | Sprout Social / Hootsuite Insights | Integrated with scheduling; stronger reporting |
| Enterprise | Brandwatch | Full NLP suite; social listening + analytics |
NLP language limitation — critical for EA clients: English NLP models from all major tools are reliable for standard sentiment detection. Luganda and Swahili NLP is unreliable across all commercial tools as of 2025 — sentiment scores for non-English content will be inaccurate, sometimes inverted. Always manually review Luganda and Swahili comments regardless of tool used.
The Net Sentiment Score expresses brand health as a single number, calculated monthly and tracked as a trend. Based on methodology in Funk (2013).
Formula:
NSS = (Positive mentions − Negative mentions) ÷ Total mentions × 100
Mixed comments are excluded from the numerator. Neutral and Mixed comments are included in the denominator (total mentions).
Worked example:
Express NSS as a number (e.g., 52.1). Never express it as a word (e.g., "good" or "positive").
NSS benchmarks for EA service businesses:
| Score | Assessment |
|---|---|
| +60 or above | Strong — brand advocates outweigh detractors significantly |
| +40 to +59 | Healthy — majority positive; address recurring negatives |
| +20 to +39 | Developing — equal parts positive and neutral; negatives need attention |
| 0 to +19 | Concerning — negatives approaching parity with positives; investigate root causes |
| Below 0 | Crisis territory — negatives outweigh positives; trigger playbook-crisis-communications |
Tracking NSS over time: Report NSS as a monthly trend, not a single snapshot. A trend line is more meaningful than any individual score.
Example trend report: Month 1: +45 → Month 2: +48 → Month 3: +52 (improving, +7 over quarter)
Tip: A stable score in the +40s is a better signal than an improving score that started at +15. Always contextualise the number against the trend and the starting baseline.
Share of Voice measures the client's portion of the total brand conversation in their competitive set. Based on methodology in Schaffer (2013).
Formula:
SOV = Client brand mentions ÷ (Client + Competitor A + Competitor B + ...) × 100
Worked example:
Automated monitoring tools are ideal but not always available. Use the following methods in order of preference:
Record raw mention counts weekly and sum to a monthly total for SOV calculation.
| SOV | Assessment | Recommended Action |
|---|---|---|
| Below 25% | Low share — below competitive threshold | Increase content volume; generate more PR moments; consider UGC campaign |
| 25–40% | Competitive — in the conversation | Focus on quality and differentiation; aim to win on NSS while building SOV |
| Above 40% | Dominant — leading the conversation | Maintain quality; deepen community trust; protect position from challengers |
NSS and SOV are summary metrics. Theme extraction explains what is driving those numbers. Perform this analysis monthly alongside NSS calculation.
Positive themes reveal what the audience genuinely values about the brand — and what content to amplify.
Ask the following questions of the positive comment set:
These themes are content pillars hiding in plain sight. Use them as future content topics, social proof copy, and campaign angles. Refer to 10-content-pillars when translating positive themes into a content framework.
Negative themes reveal operational and product problems that marketing cannot solve — and should not be asked to solve.
Ask the following questions of the negative comment set:
Recurring negative themes are product and service improvement briefs, not marketing problems. Escalate to operations when a theme appears 5 or more times in a month.
Apply the same process to positive comments to produce "Top Positive Themes".
This process takes approximately 20–30 minutes per platform once the comment export is complete.
Sentiment data has no value unless it drives a decision (Funk, 2013). Every monthly sentiment report must conclude with at least one named strategic action. Use the following decision table to identify the correct response to each type of finding.
| Finding | Strategic Action |
|---|---|
| NSS declining 3 months in a row | Audit content quality and community management response times; identify whether the cause is content failure or service failure |
| One negative theme appearing 5+ times per month | Treat as a product/service improvement brief — escalate to the client's operations team; do not attempt to resolve through marketing |
| Competitor NSS significantly higher | Analyse their top-performing positive content for insight into what their audience values; apply learnings to content planning |
| SOV declining | Increase content frequency or launch a PR/UGC campaign to generate more brand-attributable mentions |
| Positive theme emerging organically | Build a content series around this theme immediately; document it as a content pillar in 10-content-pillars |
| NSS spike after a campaign | Campaign worked — document the content type, timing, and audience response; replicate in future campaigns |
| NSS drop after a campaign | Campaign content may have missed the mark — review content, tone, and targeting; debrief with client before next campaign |
| SOV above 40% with declining NSS | Volume is high but quality is suffering; reduce content frequency and invest in quality; community management may be under-resourced |
Each action must have a named owner (consultant, client, operations) and a deadline. Data without an owner and a deadline does not produce change.
Produce this report monthly. Deliver it to the client alongside the written monthly report (meta-reporting) or include it in the evidence handoff to design-system-skills when a presentation is commissioned. Do not claim a local deck route or produce this as a standalone document unless the client has requested a dedicated sentiment briefing.
MONTHLY SENTIMENT REPORT — [Client Name] — [Month, Year]
NSS this month: [score] | Last month: [score] | Trend: [improving / stable / declining]
SOV this month: [%] | Competitor A: [%] | Competitor B: [%]
Top 3 positive themes this month:
1. [Theme 1] — [X mentions]
2. [Theme 2] — [X mentions]
3. [Theme 3] — [X mentions]
Top 3 negative themes this month:
1. [Theme 1] — [X mentions] — [Product / Service / Price / Fulfilment / Brand]
2. [Theme 2] — [X mentions] — [Category]
3. [Theme 3] — [X mentions] — [Category]
NSS 3-month trend: Month 1: [score] → Month 2: [score] → Month 3: [score]
SOV 3-month trend: Month 1: [%] → Month 2: [%] → Month 3: [%]
Recommended action this month:
[One specific strategic action — what, who, by when]
Populate all fields from the NSS calculation, SOV calculation, and theme extraction outputs produced in this session. Do not leave fields blank; if data is unavailable, note the reason and the plan to collect it next month.
playbook-sentiment-listening — operational setup: tools, keyword lists, dashboards, weekly monitoring routines (produces the raw data this skill analyses)meta-reporting — monthly written performance report; integrate the sentiment report template into the reporting sectionmeta-competitor-analysis — full competitor benchmarking; SOV from this skill feeds into the competitive landscape sectionplaybook-crisis-communications — invoke immediately when NSS falls below 010-content-pillars — use positive theme extraction outputs to define or refresh content pillars