以 AI 机器人身份加入视频会议,提供语音、虚拟形象与屏幕共享四种模式。
记忆
AuditClaw Azure
试用Azure compliance evidence collection for auditclaw-grc. 12 read-only checks across storage, NSG, Key Vault, SQL, compute, App Service, and Defender for Cloud.
它能做什么
Companion skill for auditclaw-grc. Collects compliance evidence from Azure subscriptions using read-only API calls.
技能文档
AuditClaw Azure
Companion skill for auditclaw-grc. Collects compliance evidence from Azure subscriptions using read-only API calls.
12 checks | Reader + Security Reader roles only | Evidence stored in shared GRC database
Security Model
- Read-only access: Requires only Reader + Security Reader roles (subscription-level). No write/modify permissions.
- Credentials: Uses
DefaultAzureCredential(service principal env vars,az login, or managed identity). No credentials stored by this skill. - Dependencies: Azure SDK packages (all pinned in requirements.txt)
- Data flow: Check results stored as evidence in
~/.openclaw/grc/compliance.sqlitevia auditclaw-grc
Prerequisites
- Azure credentials configured (service principal or
az login) pip install -r scripts/requirements.txt- auditclaw-grc skill installed and initialized
Commands
- "Run Azure evidence sweep": Run all checks, store results in GRC database
- "Check Azure storage security": Run storage-specific checks
- "Check Azure network security": Run NSG checks
- "Check Azure Key Vault": Run Key Vault checks
- "Check Azure SQL compliance": Run SQL Server checks
- "Check Azure VM encryption": Run compute checks
- "Check Azure App Service": Run App Service checks
- "Check Azure Defender": Run Defender for Cloud checks
- "Show Azure integration health": Last sync, errors, evidence count
Usage
All evidence is stored in the shared GRC database at ~/.openclaw/grc/compliance.sqlite via the auditclaw-grc skill's db_query.py script.
To run a full evidence sweep:
python3 scripts/azure_evidence.py --db-path ~/.openclaw/grc/compliance.sqlite --all
To run specific checks:
python3 scripts/azure_evidence.py --db-path ~/.openclaw/grc/compliance.sqlite --checks storage,network,keyvault
To list available checks:
python3 scripts/azure_evidence.py --list-checks
Check Categories (7 files, 12 findings)
| Check | What It Verifies |
|---|---|
| storage | HTTPS-only transfer, TLS 1.2+, public blob access disabled, network default deny |
| network | NSG no unrestricted SSH (port 22), no unrestricted RDP (port 3389) |
| keyvault | Soft delete + purge protection enabled |
| sql | Server auditing enabled, TDE encryption on all databases |
| compute | VM disk encryption (encryption at host) |
| appservice | HTTPS-only + TLS 1.2+ |
| defender | Defender plans enabled (Standard tier) for critical resource types |
Authentication
Uses DefaultAzureCredential from azure-identity. Supports:
- Service principal:
AZURE_CLIENT_ID+AZURE_TENANT_ID+AZURE_CLIENT_SECRET - Azure CLI:
az login - Managed identity (when running in Azure)
Minimum roles: Reader + Security Reader (subscription-level)
Evidence Storage
Each check produces evidence items stored with:
source: "azure"type: "automated"control_id: Mapped to relevant SOC2/ISO/HIPAA controlsdescription: Human-readable finding summaryfile_content: JSON details of the check result
Setup Guide
When a user asks to set up Azure integration, guide them through these steps:
Step 1: Create Service Principal
az ad sp create-for-rbac --name auditclaw-scanner --role Reader --scopes /subscriptions/
Step 2: Add Security Reader Role
az role assignment create --assignee --role "Security Reader" --scope /subscriptions/
Only 2 roles needed: Reader + Security Reader (subscription-level).
Step 3: Configure Credentials
Set environment variables from the service principal output:
- AZURE_CLIENT_ID (appId)
- AZURE_CLIENT_SECRET (password)
- AZURE_TENANT_ID (tenant)
- AZURE_SUBSCRIPTION_ID
Step 4: Verify Connection
Run: python3 {baseDir}/scripts/azure_evidence.py --test-connection
The exact roles are documented in scripts/azure-roles.json. Show with:
python3 {baseDir}/../auditclaw-grc/scripts/db_query.py --action show-policy --provider azure
相关技能
把自然语言描述转为结构化 JSON,并由 mcp-diagram-generator MCP 服务生成 Draw.io、Mermaid 或 Excalidraw 图表文件。
在本地磁盘以分类纯 Markdown 文件保存需要长期留存的事实,与智能体内置记忆并存。
诊断生产力系统反复失效的根因,给出最小干预——容量测算、瓶颈定位、可靠的本地记录。
为自然搜索排名提供站点审计、内容撰写与竞品分析。
从 AdMapix API 拉取广告创意、应用、榜单和收入预估等数据,原样返回结构化 JSON。
mailnike 的更多技能
浏览全部技能基于本地 SQLite,提供复式记账 ERP,覆盖会计、开票、库存、薪资与报表全流程。
Web dashboard for OpenClaw. Browser-based UI for any installed skill. Schema-driven rendering, JWT auth, RBAC, AI chat, real-time updates. Install web dashbo...
AI-native GRC (Governance, Risk, and Compliance) for OpenClaw. 97 actions across 13 frameworks including SOC 2, ISO 27001, HIPAA, GDPR, NIST CSF, PCI DSS, CI...
AI-native ERP system with self-extending OS. Full accounting, invoicing, inventory, purchasing, tax, billing, HR, payroll, advanced accounting (ASC 606/842,...
Web dashboard for OpenClaw. Browser-based UI for any installed skill. Schema-driven rendering, JWT auth, RBAC, AI chat, real-time updates. Install web dashbo...
Identity provider compliance checks for auditclaw-grc. 8 read-only checks across Google Workspace (MFA, admin audit, inactive users, passwords) and Okta (MFA...