← Back to Skills Marketplace
ashley-aihr

Compensation Repo

by Ashley · GitHub ↗ · v0.5.0 · MIT-0
cross-platform ✓ Security Clean
140
Downloads
0
Stars
0
Active Installs
3
Versions
Install in OpenClaw
/install hr-compensation-checks
Description
帮 HR 做定薪判断、band 对标、市场调研摘要,以及个税社保公积金申报前检查,先看值不值,再看会不会出风险。 / Help HR teams with compensation review, band and market checks, and payroll filing prechecks.
README (SKILL.md)

定薪与申报检查助手 / Compensation Decision Assistant

当用户在处理两类薪酬工作时使用这个 skill:

  1. 定薪判断:band、市场调研、内部公平、offer 建议
  2. 申报检查:个税、社保、公积金申报前排雷

目标不是手算工资,而是输出:

  1. 结论
  2. 依据
  3. 风险
  4. 待办
  5. 可直接发给内部协作方的说明

如果用户第一次使用或输入很乱,先读 references/real-user-scenario.md。 如果需要工作流背景,读 references/compensation-workflows.md。 如果需要最新政策、城市口径和系统操作依据,读 references/china-compensation-policy-kb-2026.md。 如果需要理解动态市场数据怎么分层、哪些能当正式依据,读 references/dynamic-market-data-architecture.md

路由规则

根据输入内容路由到下面动作之一:

  1. review_compensation_band_and_offer 触发条件:输入里有 band、市场分位、候选人期望、内部参考、预算中的任意组合。
  2. precheck_payroll_filing 触发条件:输入里有个税、社保、公积金申报字段,或月度申报名单、员工状态、主体信息。

如果用户不知道该选哪个动作:

  1. 有申报名单、基数、主体、缴纳地,就走 precheck_payroll_filing
  2. 有 band、市场分位、候选人期望,就走 review_compensation_band_and_offer

review_compensation_band_and_offer,必须区分:

  1. official_policy
  2. public_market_signal
  3. paid_survey_data
  4. internal_company_data

如果只有 public_market_signal,不允许把结论写成正式定薪建议。

输出协议

处理任意薪酬场景时,始终输出:

normalized_data
decision_summary
decision_basis
missing_information
risk_summary
priority_issues
next_action
message_draft
record_update
human_confirmation_needed
compliance_warning_if_any

要求:

  1. decision_summary 必须先回答“怎么定”或“能不能报”。
  2. decision_basis 必须把 band、市场、内部参考或申报依据讲清楚。
  3. missing_information 只写真正影响判断或申报的缺口。
  4. risk_summary 优先写申报失败风险、内部公平风险、预算风险。
  5. priority_issues 必须按高、中、低排序。
  6. next_action 必须是 HR 今天能做的动作。
  7. message_draft 默认写给业务负责人、薪酬同事或数据提供方。
  8. human_confirmation_needed 必须写清楚还要谁确认什么。
  9. 对定薪场景,必须标明本次结论属于 正式建议弱建议 还是 仅市场信号判断

动作要求

review_compensation_band_and_offer

至少抽取:

job_family
job_level
band_min
band_mid
band_max
market_p25
market_p50
market_p75
candidate_current_pay
candidate_expected_pay
internal_peer_reference
budget_range

并优先识别:

official_policy
public_market_signal
paid_survey_data
internal_company_data
candidate_total_comp_context

结果优先顺序:

  1. 建议怎么定
  2. 为什么这么定
  3. 内部公平或预算风险
  4. 怎么和业务解释
  5. 还需要谁确认

判断规则:

  1. 同时具备 internal_company_data + paid_survey_data + candidate_current_pay_or_total_comp + budget_range 时,才可给 正式建议
  2. 只有 public_market_signal 时,只能给 市场信号判断
  3. 缺少 bandinternal_company_data 时,不得假装能完成内部公平判断
  4. 缺少 budget_range 时,不得假装能完成审批级建议
  5. 缺少 candidate_current_pay 或总包口径时,要主动降低结论强度

如果需要文件产出,运行:

node scripts/generate_band_offer_packet.js \x3Cinput.json> \x3Coutput-dir>

示例输入: assets/band-offer-review-input.sample.json 动态分层示例输入: assets/band-offer-review-input.dynamic.sample.json

precheck_payroll_filing

至少抽取:

employee_name
employee_status
legal_entity
work_city
filing_city
bank_account_status
id_number_status
taxable_income
social_base
housing_fund_base
special_deduction_status

结果优先顺序:

  1. 能不能直接报
  2. 高风险问题
  3. 按人列出的缺口
  4. 今天先处理什么
  5. 给内部同事的追回或提醒话术

如果需要文件产出,运行:

node scripts/generate_payroll_precheck_packet.js \x3Cinput.json> \x3Coutput-dir>

示例输入: assets/payroll-precheck-input.sample.json

工作原则

  1. 先给结论,再给依据,再给待办。
  2. 输入默认不干净,先归一化,不要要求用户先自己整理完。
  3. 申报检查优先抓“漏人、错主体、错城市、错基数、缺字段”。
  4. 定薪判断优先看 band、市场和内部公平,不要只盯一个数字。
  5. 缺政策口径或核心字段时,不要装得很确定,要明确降置信度。
  6. 不自动给法律结论,但要明确提示合规风险。
  7. 对公网职位薪资,只能当作市场信号,不能冒充正式薪酬调研。
Usage Guidance
Treat this as an incomplete review rather than a clean security assurance; the artifacts should be inspected successfully before installation.
Capability Assessment
Purpose & Capability
No supplied artifact evidence showed a purpose-capability mismatch.
Instruction Scope
No supplied artifact evidence showed hidden or unsafe runtime instructions.
Install Mechanism
No supplied artifact evidence showed unsafe install behavior.
Credentials
No supplied artifact evidence showed disproportionate environment access.
Persistence & Privilege
No supplied artifact evidence showed persistence or privilege abuse.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install hr-compensation-checks
  3. After installation, invoke the skill by name or use /hr-compensation-checks
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v0.5.0
**Summary:** Introduces dynamic market data handling, clearer output protocols, and improved decision logic for compensation and payroll checks. - Added support for dynamic market data input and output with new sample assets. - Refined workflow routing based on input content; now distinguishes between official, market, survey, and internal data sources. - Updated output structure to require explicit decision summaries, bases, risk levels, and human confirmation steps. - Improved clarity of documentation and protocols, including updated SKILL.md with new requirements and principles. - Expanded references and sample scenarios, including latest China compensation policy and market data architecture guides.
v0.2.0
Add a second production-ready scenario for compensation band review, market benchmark summary, and offer pricing recommendation, alongside payroll filing prechecks.
v0.1.0
Initial release: a China-focused payroll filing precheck skill for IIT, social insurance, and housing fund submission risk checks.
Metadata
Slug hr-compensation-checks
Version 0.5.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 3
Frequently Asked Questions

What is Compensation Repo?

帮 HR 做定薪判断、band 对标、市场调研摘要,以及个税社保公积金申报前检查,先看值不值,再看会不会出风险。 / Help HR teams with compensation review, band and market checks, and payroll filing prechecks. It is an AI Agent Skill for Claude Code / OpenClaw, with 140 downloads so far.

How do I install Compensation Repo?

Run "/install hr-compensation-checks" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Compensation Repo free?

Yes, Compensation Repo is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Compensation Repo support?

Compensation Repo is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Compensation Repo?

It is built and maintained by Ashley (@ashley-aihr); the current version is v0.5.0.

💬 Comments