跳到正文
cccskills
免费在 GitHub 公开

compliance-checker

Audits a codebase or business process for regulatory compliance across GDPR, HIPAA, SOC2, CCPA, and PCI-DSS. Scans for PII handling, data retention, encryption, access controls, audit logging, consent management, and data transfer issues. Generates a structured compliance report with findings, gap analysis, remediation steps, and evidence requirements.

中文简介正在准备中。当前显示原文说明。

实测评分 未实测

查看安装方法

包含的文件(6)

  • SKILL.md5.3 KB
  • references/classification.md1.9 KB
  • references/cross-framework-mapping.md3.2 KB
  • references/frameworks.md3.4 KB
  • references/output-template.md24.8 KB
  • references/scan-patterns.md10.6 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Compliance Checker

Perform thorough regulatory compliance audits against one or more frameworks, identify gaps, and produce actionable remediation guidance with evidence requirements suitable for certification preparation.

Contents

  • references/frameworks.md — the five supported frameworks (GDPR, HIPAA, SOC 2, CCPA, PCI-DSS): scope, key articles/requirements, penalties, and reference links.
  • references/scan-patterns.md — the seven scan categories, their search patterns, scan execution rules, and common pitfalls to always check.
  • references/classification.md — compliance status values, risk severity rubric, and the remediation detail fields.
  • references/cross-framework-mapping.md — control-area-to-framework mapping and the glossary.
  • references/output-template.md — the full compliance-report.md structure, including per-framework gap-analysis tables and evidence packs.

Workflow

Run the full methodology for every audit. Do not skip steps.

Phase 1: Discovery and Scoping

  1. Identify the target: codebase, infrastructure configuration, business process documentation, or a combination.
  2. Determine applicable frameworks from the data types processed, geographic scope, industry, and business model. See references/frameworks.md. When none are specified, assess all five and note applicability.
  3. Map the data flow: where regulated data enters, how it is processed, where it is stored, and how it exits.
  4. Inventory data categories present (PII, PHI, CHD, sensitive PI, special category data).

Phase 2: Scanning and Evidence Collection

  1. Scan each of the seven categories — PII/sensitive data, data retention, encryption, access controls, audit logging, consent management, and data transfer. Apply the search patterns and execution rules in references/scan-patterns.md. Record exact file paths, line numbers, and snippets for both positive findings and gaps.

Phase 3: Gap Analysis

  1. Map each finding to specific framework requirements, assign a compliance status, and assign a risk severity to every non-compliant or partial finding. Use the rubric in references/classification.md.

Phase 4: Remediation Planning

  1. For each non-compliant or partial finding, produce the remediation detail (description, regulatory reference, current state, required state, steps, evidence, effort, dependencies) per references/classification.md. Cross-reference references/cross-framework-mapping.md to flag when one action closes gaps across multiple frameworks.

Phase 5: Report Generation

  1. Generate compliance-report.md in the project root (or a user-specified location), following the mandatory structure in references/output-template.md. Every section is required; if a section has no findings, state that explicitly.

Behavioral Rules

  1. Never fabricate findings. Back every finding with scan evidence. When evidence is absent, mark UNABLE TO ASSESS rather than guess.
  2. Be specific. Reference exact file paths, line numbers, function names, and configuration keys.
  3. Distinguish technical from organizational controls. Note when a finding requires verifying organizational controls (training, policies, physical security) that code review alone cannot confirm.
  4. Consider the full stack: application code, database schemas, infrastructure configuration, CI/CD pipelines, dependency manifests, documentation, and environment configuration.
  5. Prioritize actionable findings. Specify exact changes; avoid vague recommendations like "improve security".
  6. Respect the scope. Audit only what is provided. Do not assume broader infrastructure without evidence.
  7. Account for frameworks and libraries that handle compliance concerns (ORM parameterized queries, framework CSRF protection); credit them as positive findings but verify configuration.
  8. Avoid false positives. Missing a marginal finding is better than reporting a non-gap. Err toward accuracy.
  9. Cross-reference across frameworks. Note when a single remediation addresses multiple regulations.
  10. Date and version everything so the report can be compared against future audits.

Responding to Requests

  • "Audit this codebase" / "check compliance": Confirm which frameworks to focus on (or assess all five), run Phases 1-5, generate compliance-report.md, and summarize the most critical findings.
  • "Check for [framework]": Focus on that framework, still scan all seven categories, and map findings only to it.
  • "Check [category]" (e.g., "check encryption"): Focus the scan on that category and map findings to all applicable frameworks.
  • "Are we [requirement] compliant?" (e.g., GDPR Art. 17): Focus on the relevant controls, give a direct assessment with evidence, and offer remediation if gaps exist.
  • Business process documentation instead of code: Adapt the methodology to document analysis, focus on policy and procedure requirements, and note that technical controls cannot be verified without codebase access.

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Audit websites for accessibility issues and WCAG compliance. Use when checking accessibility, fixing a11y issues, or ensuring WCAG compliance.

中文简介正在准备中。当前显示原文说明。

OneWave-AI/claude-skills3682026年10月20日 更新

Deploy a 2-layer parallel agent hierarchy for large, parallelizable work — big refactors, multi-file migrations, codebase-wide audits, bulk generation. A top-tier commander (Fable or Opus) orchestrates the swarms; the user picks a power level (Max Power / Heavy / Balanced / Economy) that sets the Opus/Sonnet/Haiku model mix per layer. Layer 1 is 3-50+ specialist agents, each with its own full context window; Layer 2 is 2+ sub-agents per member. Includes git safety, tiered sizing, a pre-deploy gate, phantom-completion checks, and multi-wave follow-up.

中文简介正在准备中。当前显示原文说明。

OneWave-AI/claude-skills3682026年10月20日 更新

Use to test whether an AI agent, prompt, or skill actually works before or after it ships — build a small eval set, score runs, and catch regressions when the prompt or model changes. Trigger when the user asks whether a prompt is good, wants to compare models or versions, sees inconsistent agent output, or is about to put an agent in front of customers.

中文简介正在准备中。当前显示原文说明。

OneWave-AI/claude-skills3682026年10月20日 更新

Deploys swarms of sub-agents for massive parallel data processing tasks. Unlike agent-army (which is for code changes), this is for DATA tasks -- processing 1000 documents, analyzing datasets, bulk content generation. Configurable swarm size, task distribution, result aggregation, progress tracking, and error recovery.

中文简介正在准备中。当前显示原文说明。

OneWave-AI/claude-skills3682026年10月20日 更新

Designs and deploys custom agent teams for specific business workflows. Interactive discovery of business processes, then generates complete team configurations with specialized agent roles, tool access, communication protocols, and handoff rules.

中文简介正在准备中。当前显示原文说明。

OneWave-AI/claude-skills3682026年10月20日 更新

Agent-to-Agent (A2A) communication protocol. Connect two or more Claude agents that pass messages, share context, delegate tasks, and collaborate. Implements structured handoffs, shared memory, and multi-agent conversations.

中文简介正在准备中。当前显示原文说明。

OneWave-AI/claude-skills3682026年10月20日 更新

查看 OneWave-AI 的全部技能

报告此技能的问题