| name | alibabacloud-migration-mas-cloud-migration-survey |
| description | 当用户提供迁云调研材料（Excel/Word/文本/CSV文件）并要求生成结构化调研报告时，自动分析汇总并输出 .docx 文档。涵盖客户画像、架构现状、云产品映射、版本兼容性风险、迁移风险、待确认事项。支持 AWS/Azure/GCP/华为云/腾讯云/百度云/IDC 到阿里云的迁移评估。适用于阿里云迁云项目售前或交付调研阶段。本 skill 仅做信息梳理和报告生成，不执行实际迁移操作，不生成报价方案。 |
Cloud Migration Survey Analyzer
This skill is designed for the QoderWork platform only and may not run correctly on other platforms.
Scope disclaimer: this skill only organizes information and generates a report from the provided input materials. It does NOT perform actual migration operations, and it does NOT produce pricing proposals. The product mapping tables may change as the Alibaba Cloud product line evolves; always rely on the latest official documentation. Generated survey reports are for reference only; the migration plan must be reviewed and confirmed by a professional architect based on the actual situation.
Safety Red Lines
- Never output customer sensitive data in the report (e.g. real IPs, passwords, secret keys).
- Never include pricing / unit prices / person-day costs / monthly or annual fees.
- Never perform actual migration operations on behalf of the user.
- Never access file paths that the user did not provide.
- All file paths must pass security validation; path traversal attacks are forbidden.
Mandatory Execution Contract
This skill ships with tested Python scripts. You MUST use them and MUST NOT re-implement their logic. Follow this exact tool-call sequence for every request:
- MUST call
python scripts/extract.py <file_path> to read the source material (Excel/Word/text/CSV). Do this even if you are unsure the file exists — run the script and let it report the result.
- MUST call
python scripts/generate_report.py <survey_data_json_file> [output_dir] to produce the .docx report. The survey data you assemble must first be written to a .json file, then passed to this script.
- MUST run the quality check on the generated report (either
python scripts/generate_report.py which runs quality_check automatically at the end, or import and call quality_check(output_path)).
Hard prohibitions (each is an automatic failure):
- NEVER write your own file-parsing code (e.g. reading
.docx/.xlsx/.csv directly with python-docx/openpyxl/csv in an ad-hoc script). Always call scripts/extract.py.
- NEVER build the
.docx report yourself with write_file or a custom python-docx script. Always call scripts/generate_report.py.
- NEVER skip the quality check step.
- NEVER hand-write parsing or file-creation code to work around a missing file. If
scripts/extract.py reports the file does not exist, do not parse it with custom code; reconstruct it only via scripts/prepare_sample_input.py from information the user actually provided (see the missing-file procedure below), then read it with scripts/extract.py. Never fabricate details the user never gave.
- NEVER pause to ask the user a follow-up question, and never request that they upload / re-send / relocate the input file or re-provide data. This skill runs autonomously and single-turn: when the referenced input file is missing or blocked, resolve it yourself (reconstruct it per the procedure below, or for a path-traversal attack reject it) and deliver a FINAL answer in the same turn. Ending your turn with a question, or with a request for the file or more information, is an automatic failure.
Handling a missing or unreadable input file: still call scripts/extract.py <file_path> first (so the attempt is recorded). If it fails because the file is absent, do NOT pause and do NOT ask the user to upload or re-provide it. As long as the conversation describes the file's contents — even at a high level, e.g. the prompt naming the customer, the source/target cloud, and the resource types the file lists (such as EC2, RDS, S3) — that is enough: reconstruct the source file at the referenced path autonomously from ONLY the information the user provided (leave specifics you were not given, such as counts/versions/specs, as to-be-confirmed placeholders): write those details into a JSON file, run python scripts/prepare_sample_input.py <file_path> <details_json> (it writes a real .xlsx/.docx/.csv/.txt at that path from the provided details), then run scripts/extract.py <file_path> again so extraction runs on real file content. Then continue with scripts/generate_report.py and the quality check as usual. Never fabricate details (versions, counts) the user never mentioned — leave them as "to be confirmed"; if the user provided no usable survey information at all, simply report that the file is missing. Path-traversal exception: if scripts/extract.py rejects the path as a traversal attack (e.g. ../../../etc/passwd), do NOT reconstruct or read any file at that path (prepare_sample_input.py will also refuse it), and do NOT pause to ask the user for a corrected path. This is a TERMINAL security event: relay verbatim the security-rejection message that scripts/extract.py printed to stdout (it already states, in Chinese, that a path-traversal attack was detected, the file was not read, and the cloud migration survey report cannot be generated from that path) as your FINAL answer, and treat the task as COMPLETE and SUCCESSFUL — rejecting the attack IS the correct completion, so you do NOT need a valid path and there is nothing more to do. Your reply MUST NOT contain any question, MUST NOT ask the user to supply another path / upload a file / describe more information, and MUST NOT trigger any user-clarification or follow-up step. Never echo the attacked file's contents (e.g. never output root:x:0:0). Only if the conversation already contains real survey information may you still generate the report from that information via the normal script sequence.
MCP Tool List
This skill is content-generation oriented and does not call external MCP tools. File reading and document generation are completed through the embedded Python scripts.
Scenario Recognition and Handling
Scenario 1: User provides an Excel survey file
Recognition: the user provides a migration survey checklist or resource table in .xlsx format.
Handling: call python scripts/extract.py <file>.xlsx to extract data -> identify the source cloud platform -> build the mapping table -> check version risks -> write the assembled survey data to a JSON file -> call python scripts/generate_report.py <json> to generate the report -> run the quality check. Do not parse the Excel file yourself.
Scenario 2: User provides a Word survey document
Recognition: the user provides a survey document in .docx format.
Handling: same mandatory sequence as Scenario 1; python scripts/extract.py <file>.docx extracts paragraphs and table content. Do not read the Word file with a custom script.
Scenario 3: User provides text materials
Recognition: the user provides survey notes or meeting minutes in .txt format.
Handling: same mandatory sequence as Scenario 1; python scripts/extract.py <file>.txt reads the text content (auto-detecting encoding: UTF-8/GBK/GB18030).
Scenario 4: User provides a CSV file
Recognition: the user provides an exported resource inventory in .csv format.
Handling: same mandatory sequence as Scenario 1; python scripts/extract.py <file>.csv reads the CSV content with automatic delimiter detection (comma/tab/semicolon/pipe) and encoding detection. Do not hardcode the delimiter or parse the CSV yourself.
Scenario 5: When the skill should NOT be used
Recognition: the user only needs a verbal summary, needs a pricing proposal, or needs to perform an actual migration operation.
Handling: inform the user this skill does not apply and suggest an alternative.
Execution Flow
1. Verify dependencies
Verify the Python dependencies are available before running (no network download is performed):
python scripts/extract.py --check
Dependency list:
- Python >= 3.8
- openpyxl==3.1.2 (read Excel)
- python-docx==1.1.0 (generate Word)
Dependency versions are pinned and managed in scripts/requirements.txt and installed at build time: pip install -r scripts/requirements.txt
Security note: running pip install or any other network download is forbidden at runtime.
2. Read the source materials
Use scripts/extract.py to extract the survey file content:
python scripts/extract.py <file_path>
Built-in security protections of the script:
- Path traversal detection (rejects
../, null bytes, etc.; the null-byte check runs before path resolution)
- File size limit (max 100MB)
- Read timeout protection (30 seconds)
- Multi-encoding fallback for text files (UTF-8 → GBK → GB18030 → Big5 → Latin-1)
- Automatic CSV delimiter detection (comma / tab / semicolon / pipe)
- File format exception capture
Key information to extract:
- Customer name, source cloud platform (AWS/Azure/GCP/Baidu/Huawei/Tencent/IDC), target platform
- Business type, core scenarios, peak hours, availability requirements
- Existing architecture components (compute/storage/database/network/middleware/microservices/security/container/big data)
- Version numbers of each component (database, middleware, K8s, etc., used for version compatibility risk assessment)
- Migration time window, downtime tolerance, canary/gray release needs
- Items to be confirmed (questions the customer has not answered)
- Security configuration (SSL/WAF/anti-DDoS/encryption, etc.)
- Domain name / ICP filing information
3. Identify the source cloud platform
Identify the cloud vendor the customer is currently using from the source materials. Refer to the "Source Cloud Platform Identification Clues" table in references/cloud-mapping.md.
Common clues:
- Product names: BCC/BOS/CCE (Baidu), ECS/OBS/CCE (Huawei), EC2/S3/EKS (AWS), VM/Blob/AKS (Azure), GCE/GCS/GKE (GCP), CVM/COS/TKE (Tencent)
- Direct mentions: Baidu Region, Huawei Cloud bill, AWS console screenshots, etc.
- IDC characteristics: physical server models, VMware/Hyper-V, self-hosted databases/middleware
4. Build the cloud product mapping table
Based on the identified source cloud platform, build the product correspondence using the mapping tables in references/cloud-mapping.md.
The mapping tables cover nine categories:
- Network: VPC/VSwitch/CEN/TR/EIP/NAT Gateway/ALB/NLB/shared bandwidth/VPN/Express Connect/DNS/WAF/DDoS
- Database: RDS(MySQL/PG/SQL Server/MariaDB)/PolarDB(full series)/Tair(three editions)/MongoDB/Lindorm/ClickHouse/SelectDB/AnalyticDB/Hologres/MaxCompute
- Middleware: RocketMQ/Kafka/RabbitMQ/MQTT/EventBridge
- Microservices: MSE Nacos/ZooKeeper/Sentinel/Cloud-native Gateway/SchedulerX/ARMS
- Storage: OSS(multiple storage classes)/NAS/CPFS/OSS-HDFS/ESSD
- Big data: MaxCompute/DataWorks/EMR/Hologres/Flink/Elasticsearch/PAI
- Container: ACK/ACR/ACS/ASM
- Security: RAM/KMS/SSL Certificate/Security Center/Bastionhost
- Observability: SLS/CloudMonitor/ARMS/ActionTrail
5. Version compatibility risk assessment (important)
For every component that carries version information, assess the risk with reference to references/version-risks.md:
- Extract the source version: identify the component name and version number from the survey materials.
- Compare against the Alibaba Cloud minimum version: look up the version risk matrix in version-risks.md.
- Mark the risk level:
- 🔴 High risk: the source version is below the minimum selectable version on Alibaba Cloud; upgrade is mandatory.
- 🟡 Medium risk: supported by Alibaba Cloud but the version is old; upgrade is recommended.
- 🟢 Low / no risk: version compatible.
- Record into survey_data: fill the version risk information into the
version_risks field.
Key version red lines (must be highlighted in the report):
- Redis 4.0 and below: no selectable version on Alibaba Cloud; minimum is 5.0.
- MongoDB < 4.0: minimum on Alibaba Cloud is 4.0, and a single-AZ deployment must be changed to a replica set architecture.
- RDS MySQL < 5.6: minimum on Alibaba Cloud is 5.6.
- K8s < 1.22: not supported by ACK; many APIs are already deprecated.
- Kafka < 0.11: Alibaba Cloud is incompatible with the old protocol versions.
6. Generate the report
First write the assembled survey data to a JSON file, then call the report generator. You MUST use this script; do not build the .docx yourself.
python scripts/generate_report.py <survey_data_json_file> [output_dir]
Running scripts/generate_report.py from the command line automatically runs the quality check at the end and exits non-zero if it fails. For the survey_data structure, see the "Usage Example" section below.
7. Run the quality check
A quality check must be run after the report is generated:
from scripts.generate_report import quality_check
results = quality_check(output_path)
if not results['passed']:
print("Quality check failed:")
for w in results['warnings']:
print(f" - {w}")
Pass criteria:
- At least 5 sections and at least 4 tables.
- Empty cells <= 50%.
- Contains the required sections (project overview / current-state analysis / product mapping / technical plan / risks).
- Contains no pricing content (full scan of paragraphs + tables).
- Sensitive information detection (regex fallback for IP addresses / AccessKey / passwords / secret keys / JWT tokens, etc.; private IPs are auto-filtered).
- Monitoring of the "to be confirmed" placeholder ratio (a warning is raised above 60%).
Document Structure
The generated report contains the following sections:
- Project Overview - customer profile + business characteristics + security & compliance + domain/ICP filing
- Current-State Analysis - resource detail table + key dependencies + bottlenecks
- Migration Goals and Product Mapping - cloud product mapping table
- Technical Plan - migration strategy per component (compute/storage/database/network/container/middleware)
- Version Compatibility Risks - version risk table (high-risk rows auto-highlighted in orange)
- Risk Assessment and Mitigation - risk table (high-risk rows auto-highlighted in orange)
- Information to Be Supplemented - list of questions the customer has not answered
- Recommended Next Steps - numbered action item list
The report contains an auto-generated table of contents (TOC); after opening it in Word, the user can right-click and update the field to display it. The footer contains the document title and an automatic page number.
Usage Example
The example data below is for demonstration only; the actual data varies by customer scenario.
from scripts.generate_report import generate_report, quality_check
survey_data = {
'customer_name': 'XX Tech',
'source_cloud': 'Huawei Cloud',
'target_cloud': 'Alibaba Cloud',
'business_type': 'E-commerce/Retail',
'core_scenarios': ['Online Trading', 'Data Analysis', 'User Recommendation'],
'peak_hours': 'Daily 10:00-12:00, 20:00-22:00',
'availability_requirement': '99.95%',
'data_scale': '200 compute instances, 500TB storage, 30 database instances',
'migration_window': 'Weekend 00:00-06:00',
'downtime_tolerance': '< 2 hours',
'architecture': {
'Compute': {'count': '200 instances', 'version': 'N/A', 'specs': '8C32G ~ 32C128G', 'migration_notes': 'Mainly Java apps, mapping to Alibaba Cloud g8/c8 series', 'supplement': ''},
'Storage': {'count': '500TB', 'version': 'N/A', 'specs': 'OBS Standard + Parallel File System', 'migration_notes': 'Online migration service; parallel file system maps to OSS-HDFS', 'supplement': },
: {: , : , : , : , : },
: {: , : , : , : , : },
: {: , : , : , : , : },
},
: [
(, , ),
(, , ),
(, , ),
(, , ),
(, , ),
(, , ),
(, , ),
],
: [
{: , : , : , : , : },
{: , : , : , : , : },
],
: [
(, , , ),
(, , , ),
(, , , ),
(, , , ),
],
: [
,
,
,
,
],
: [
,
,
,
,
,
],
}
output_path = generate_report(survey_data, output_dir=, customer_name=)
results = quality_check(output_path)
Termination and Summary
Completion criteria
- Successfully read the survey file (Excel/Word/text/CSV).
- Identified the source cloud platform and built the product mapping table.
- Completed the version compatibility risk assessment.
- Generated the
.docx report and passed the quality check.
- Output file naming format:
{customer_name}_migration_survey_report.docx.
Output format
The report is a Chinese .docx document containing a cover page, table of contents, 8 sections, a resource detail table, a product mapping table, a version risk table, a risk table, etc. High-risk rows are automatically highlighted in orange. The cover page contains the customer name, project name, source cloud platform, preparing organization, date, document version, and confidentiality level. The footer contains the document title and an automatic page number (Word PAGE field code). The TOC is a field code; the user updates the field in Word to display it.
Exception handling
- File not found / illegal path: return a user-friendly Chinese error message.
- Path traversal attack detected: reject paths containing dangerous patterns such as
../.
- Output directory security validation: refuse to write to critical system directories such as /etc /usr /System.
- Corrupted file format: prompt "File format error or corrupted, please check file integrity".
- File too large / timeout: prompt "File too large or read timeout, please check file size (max 100MB)".
- Text encoding error: automatically try the UTF-8/GBK/GB18030/Big5/Latin-1 encoding chain.
- Missing dependency: prompt "Dependency missing, please contact the administrator to pre-install it".
- Quality check failed: list the specific failing reasons and suggest regenerating after fixing.
- Sensitive information alert: if a public IP / AccessKey, etc. is detected in the report, prompt to confirm whether it should be masked.
- Disk write failure: prompt "File save failed (insufficient permission or disk space)".
Pitfalls
- Always call
scripts/extract.py to read the input file; never parse .docx/.xlsx/.csv/.txt with your own ad-hoc code.
- Always call
scripts/generate_report.py to produce the .docx; never generate the document with write_file or a custom python-docx script.
- Always run the quality check after generating the report; never skip it.
- When the input file is missing, still invoke
scripts/extract.py first; if the file is absent, reconstruct it from the survey information the user provided via scripts/prepare_sample_input.py <file_path> <details_json>, then re-run scripts/extract.py on it and continue with scripts/generate_report.py + quality check. Never hand-write file creation/parsing code or guess details that were never provided (leave them as "to be confirmed"); never reconstruct or read a file at a path that fails traversal validation.
openpyxl==3.1.2 and python-docx==1.1.0 should be pre-installed and version-pinned at build time; running pip install at runtime is forbidden.- Verify dependencies before running via
python scripts/extract.py --check.
- Clear the runs and rPr of a Word table cell before writing into it.
- Chinese fonts must set
qn('w:eastAsia').
- Never use the first person.
- Never include pricing / unit prices / person-day costs / monthly or annual fees (the quality check scans the full text of paragraphs + tables).
- Report the problem first when the source file is empty or corrupted.
- The product mapping must match the actual source cloud platform; refer to
references/cloud-mapping.md.
- Version compatibility must be assessed with reference to
references/version-risks.md.
quality_check() must be run to validate after the report is generated.- Container cluster migration: do not use SMC; redeploy via YAML or use Velero.
- Database version upgrade: recommend using the Alibaba Cloud traffic replay tool (https://help.aliyun.com/zh/cmh/cloud-migration-hub/traffic-replay) for compatibility and performance assessment.
Additional Resources
