蚂蚁社保深圳社保计算器

概述

Ant Shenzhen Social Security Calc

Overview

Calculate Shenzhen enterprise-employee 五险 (pension / medical / unemployment / work-injury / maternity) contributions from a single input base, then render a clean branded PNG card with the 蚂蚁社保 logo pinned to the bottom-right. Numbers come from the 2025-07 policy cycle (valid 2025-07-01 ~ 2026-06-30): pension 16%/8%, medical 一档 5%/2%, unemployment 0.8%/0.2%, work-injury 0.2%~1.4% (industry-tier, defaults to lowest), maternity 0.5%/0%. Only Shenzhen; other cities are out of scope.

When to use this skill

Trigger when the user wants:

• A Shenzhen 五险 contribution breakdown (个人/企业/合计) for a given 缴费基数.
• A PNG / image card to embed in 小红书 / 公众号 / 客户答疑 / HR 培训 materials.
• A quick "公司帮我交了多少 / 我自己交了多少" estimate for Shenzhen employees.

Do NOT trigger for: 广州 / 杭州 / 上海 / 北京 (use the sibling skill in that folder), 公积金 (out of scope), or non-Shenzhen tax/个税 questions.

Workflow

1. Confirm the city is 深圳. If the user only says "社保" without a city, ask which city — there are sibling skills (ant-guangzhou-..., ant-hangzhou-..., beijing-..., shanghai-...).
2. Extract the 缴费基数. Accept any of:
   • A pure number (e.g. 15000, 1.5w → 15000).
   • A range or typical value (e.g. "按 10000 算") — take the midpoint or user's stated value.
   • If the user gives a salary rather than a 基数, ask once: "基数一般等于你的月工资总额，按 X 算吗？" Default to confirming.
3. Pick a 工伤 tier unless the user specified one.
   • Default = lowest 0.2% (most office workers).
   • If user mentions 行业 / 风险 (e.g. 建筑 / 制造业 / 物流), bump to 0.4%~1.4% accordingly and pass --injury 0.0XX (see references/rates.md for the 8-tier table).
   • If user says 深户, the standard skill ignores the extra 1% 地方补充; mention it in the response if relevant.
4. Run the calculation + render:

   /Users/corey/.workbuddy/binaries/python/versions/3.13.12/bin/python3 
     ~/.workbuddy/skills/ant-shenzhen-social-security-calc/scripts/calc_shenzhen_ss.py 
     <BASE> [--injury 0.002~0.014] --image /tmp/shenzhen_ss_<BASE>.png

The script auto-clamps each 险种 to its own [min, max] range. It prints a friendly text table to stdout AND writes the PNG.

1. Present to the user:
   • Show the PNG with present_files (mandatory — this is a visual deliverable).
   • Quote the headline numbers (个人合计 / 企业合计 / 总计) in 1~2 lines.
   • If any 险种 got clamped (基数 < 下限 or > 上限), surface that explicitly.
   • If the user asked for JSON, run again with --json and paste the JSON.

Quick reference: rates & bases (full source in references/rates.json)

合计比例 ≈ 个人 10.2% / 单位 22.5%（工伤按 0.2% 最低档时）。

Output card layout

The PNG produced by render_card.py:

• 顶部：标题"深圳 · 社保缴费明细" + 输入基数 + 范围/数据版本/生效期三行小字。
• 中部表格：6 列（险种 / 实际基数 / 个人比例 / 个人金额 / 单位比例 / 单位金额）；个人金额蓝色、单位金额橙色；奇数行有浅绿底。
• 下部分块：左=个人缴纳合计（蓝），中=企业缴纳合计（橙），右=个人+企业总计（深绿），下方副标写各自合计比例。
• 底部：左侧"数据仅供参考，实际以深圳税务局/医保局核定为准"；右下角贴 assets/logo.png（蚂蚁社保 logo）。

Examples

例 1 — 标准月薪 15000（默认最低工伤档）

python3 scripts/calc_shenzhen_ss.py 15000 --image /tmp/shenzhen_ss_15000.png

个人 ¥1,530.00 | 企业 ¥3,375.00 | 总计 ¥4,905.00 / 月

例 2 — 建筑行业基数 20000，工伤 1.0%

python3 scripts/calc_shenzhen_ss.py 20000 --injury 0.010 --image /tmp/shenzhen_ss_20000_injury.png

例 3 — 基数过低 3000，触发医疗/养老 clamp

python3 scripts/calc_shenzhen_ss.py 3000 --image /tmp/shenzhen_ss_low.png

养老实际基数=4775（已 clamp），医保实际基数=6733（已 clamp），失业/工伤=3000（>2520 未 clamp）。

例 4 — 仅取 JSON 数据

python3 scripts/calc_shenzhen_ss.py 15000 --json

Updating the data

When the 2026-07 cycle publishes new limits/rates, edit only:

• references/rates.json — base / bases_by_item / rates / meta.valid_period / meta.updated
• references/rates.md — human-readable mirror

No code change needed. Bump agent_created skill version via SkillManage if you also revise the script logic.

Resources

scripts/

• calc_shenzhen_ss.py — entry point. CLI: python3 calc_shenzhen_ss.py <BASE> [--injury RATE] [--json] [--image PATH]. Returns a JSON-able dict; can also print a friendly text table.
• render_card.py — imported by calc_shenzhen_ss.py when --image is passed. Standalone usage: python3 render_card.py (reads JSON from stdin).

references/

• rates.json — authoritative rates & bases (script reads this).
• rates.md — human-readable mirror with policy notes.

assets/

• logo.png — 蚂蚁社保 品牌 logo (transparency-friendly PNG). render_card.py pastes it at the bottom-right of the card. Replace this file to update the brand mark.