Chinadaily Overseas Report Writer

Turn one uploaded Excel monthly ledger into a client-ready overseas social media monitoring report.

Workflow

Follow these steps in order.

1. Run scripts/extract_social_metrics.py <excel_path> <output_dir>.
2. Read references/writing-rules.md for scope, tone, and non-inferable items.
3. Read references/report-template.md and draft the report body using its exact section order.
4. Use the generated CSV and JSON files to fill the platform table, account table, and high-performing content case table.
5. If the user wants a Word document, place the final body into a .docx file using the available document tool in the environment. If a document tool is unavailable, still produce the complete report text and the three tables in markdown.

Input expectations

Assume the primary input is one uploaded .xlsx monthly ledger.

The workbook usually contains:

• one summary sheet named 总结
• multiple account sheets named like Facebook-Account Name
• per-post rows beginning after three header rows

Do not ask the user to restructure the workbook unless parsing fails.

Output requirements

Always produce these deliverables:

• complete report正文 in Chinese
• 平台表格
• 账号表格
• 高互动案例表

When creating the narrative, always keep this order:

1. 执行摘要
2. 本月账号矩阵总体运行情况
3. 平台层核心数据表现
4. 账号层运营表现观察
5. 内容传播表现分析
6. 高互动内容案例与传播特征总结
7. 当前运营中值得关注的问题
8. 下阶段优化建议
9. 附件说明

Chinadaily writing rules

Apply these rules every time:

• emphasize international communication logic and content insight, not only data movement
• include concrete data in every major section, but do not overload paragraphs with raw numbers
• do not claim audience age, geography, sentiment, account security, algorithm suppression, reprint volume, or comment-risk outcomes unless the workbook directly provides that evidence
• write findings as data phenomenon + content observation + professional judgment
• default to platform-account-content-cases-issues-recommendations logic
• keep tone professional, steady, and client-facing

Analytical priorities

Prefer these judgments when the workbook supports them:

• which platforms carry the main publishing burden
• which platforms have stronger single-post efficiency
• which accounts sustain the basic traffic floor
• which accounts create standout posts
• which content themes repeatedly drive interaction
• whether posting quantity and interaction efficiency are mismatched
• whether content is strongly tied to local identity or drifts into low-value generic/news content

Required references

Use these references while drafting:

• references/writing-rules.md
• references/report-template.md
• references/table-spec.md

Required script outputs

The extraction script writes:

• summary.json
• platform_summary.csv
• account_summary.csv
• top_posts.csv
• theme_samples.csv

Base the report on those files.

Fallback behavior

If some workbook fields are missing:

• still produce the report
• explicitly narrow the affected claims
• keep the structure intact
• avoid fabricating unavailable metrics

OpenClaw compatibility notes

This skill is packaged as an AgentSkills-compatible folder with SKILL.md frontmatter and supporting resources, which aligns with OpenClaw skill loading expectations. OpenClaw loads skills from workspace or managed skill folders and requires SKILL.md as the entrypoint. citeturn275372search0turn275372search1turn275372search2