Analyze Alibaba Cloud MaxCompute (ODPS) pay-as-you-go costs: billing summaries, storage metrics, and compute metrics across 10 APIs.
⚠️ MANDATORY PRODUCT CONSTRAINT:
This skill uses ONLY the 10 aliyun maxcompute CLI commands listed in API Overview below (plugin mode, version 2022-01-04).
NEVER use aliyun bssopenapi or any of its actions (billing queries, instance bills, etc.).
NEVER use other MaxCompute APIs not in the 10-API list (e.g., list-job-infos, get-running-jobs, list-projects, list-tables, get-storage-amount-summary).
NEVER use any other product CLI (aliyun odps, aliyun dataworks, aliyun compute-nest, aliyun cms, aliyun actiontrail, etc.).
If aliyun maxcompute commands fail, fix the plugin — do NOT fall back to other APIs or products.
Daily cost spike/费用突增/每日趋势 → sum-bills-by-date (MANDATORY for locating spike dates — NOT sum-daily-bills-by-item)
Paginated daily billing details/每日账单明细/按计费项的每日账单 → sum-daily-bills-by-item (MANDATORY — NOT any bssopenapi command)
Storage/存储 → sum-storage-metrics-by-type AND sum-storage-metrics-by-date (BOTH MANDATORY — sum-storage-metrics-by-type alone is NOT sufficient)
Compute jobs/计算作业/expensive jobs/贵的作业 → list-compute-metrics-by-instance (MANDATORY — NEVER use list-job-infos, get-running-jobs, get-job-resource-usage)
SQL signatures/SQL签名/重复SQL/扫描量 → list-compute-metrics-by-signature (MANDATORY for any SQL pattern analysis)
SCAN usage trends/扫描量趋势/CU usage → sum-compute-metrics-by-usage (MANDATORY — usageType=SCAN or CU; NEVER use CloudMonitor or ActionTrail)
Compute record counts → sum-compute-metrics-by-record
Prerequisites
Pre-check: Aliyun CLI >= 3.3.3 required
Run aliyun version to verify >= 3.3.3. If not installed or version too low,
run curl -fsSL https://aliyuncli.alicdn.com/setup.sh | bash to update,
or see references/cli-installation-guide.md for installation instructions.
Then [MUST] run aliyun plugin update to ensure that any existing plugins on your local machine are always up-to-date.
[MUST] CLI User-Agent — Every aliyun CLI command invocation must include:
--user-agent AlibabaCloud-Agent-Skills/alibabacloud-odps-cost-analysis
[MUST] AI-Mode Commands — When using aliyun CLI, declare AI-Mode lifecycle:
Get RegionId: Ask user which region (e.g., cn-hangzhou, cn-shanghai)
Get time range: Ask for start/end dates (convert to millisecond timestamps)
Execute: Run appropriate aliyun maxcompute CLI command with --region {REGION_ID} and --user-agent AlibabaCloud-Agent-Skills/alibabacloud-odps-cost-analysis
IMPORTANT: ALL commands MUST start with aliyun maxcompute. NEVER use aliyun bssopenapi or any other product.
Verify: Confirm results and present to user
Data Limitations
Only pay-as-you-go billing is supported
Data available from 2023-05-07 onwards
Query up to last 12 months only
Single query range: max 31 days
Costs are estimated usage-based prices (may differ slightly from actual bills)
Pre-flight Checklist (Execute BEFORE every command)
STOP-AND-CHECK RULE: Before executing EACH command, you MUST verify: (1) Does it start with aliyun maxcompute? (2) Is the API name in the 10-API list? (3) Does it include --user-agent? If ANY answer is NO, do NOT execute — fix first.
My command includes --user-agent AlibabaCloud-Agent-Skills/alibabacloud-odps-cost-analysis
My command starts with aliyun maxcompute (NOT aliyun bssopenapi, NOT any other product)
My command uses one of the 10 APIs listed in API Overview (NOT list-job-infos, get-running-jobs, list-projects, list-tables, etc.)
I have verified the maxcompute plugin is installed (aliyun maxcompute --help succeeds)
I have asked the user for RegionId (not using default)
I have the actual RegionId value from user (not placeholder)
My command includes --region {ACTUAL_REGION_ID}
Time range does not exceed 31 days
Timestamps are in milliseconds
I am NOT reading or echoing any AK/SK values
I am NOT using aliyun bssopenapi or any of its actions
If ANY check fails, STOP and fix before proceeding.
API Overview
API
Description
Method
Category
list-instances
Get instance/project list (NOT compute metrics — see list-compute-metrics-by-instance for job-level data)
GET
Instance
sum-bills
Summarize bills by project or fee item
POST
Billing
sum-bills-by-date
Daily bill trends (USE THIS for spike analysis)
POST
Billing
sum-daily-bills-by-item
Daily bill details by item (paginated drill-down)
POST
Billing
sum-storage-metrics-by-type
Storage grouped by TYPE
POST
Storage
sum-storage-metrics-by-date
Storage daily DATE trends (MUST call separately)
POST
Storage
list-compute-metrics-by-instance
Compute JOB METRICS per instance (cost, duration, input size)
POST
Compute
list-compute-metrics-by-signature
Compute by SQL signature
POST
Compute
sum-compute-metrics-by-usage
Compute usage trends
POST
Compute
sum-compute-metrics-by-record
Compute record counts
POST
Compute
⚠️ API Disambiguation — Do NOT confuse these two billing APIs:
sum-bills-by-date = Daily cost TRENDS → Use this to locate spike dates (returns cost per day)
sum-daily-bills-by-item = Daily bill DETAILS by item → Use this for drill-down after finding spikes (paginated, returns per-item breakdown)
For cost spike investigation, you MUST call sum-bills-by-date first. sum-daily-bills-by-item is optional for drill-down only.
⚠️ API Disambiguation — Do NOT confuse these two "instance" APIs:
list-instances = Returns the PROJECT/INSTANCE LIST (for scoping which projects to analyze)
These are COMPLETELY DIFFERENT APIs. list-instances does NOT return compute metrics. You MUST call BOTH.
⚠️ API Disambiguation — Do NOT confuse these two storage APIs:
sum-storage-metrics-by-type = Storage grouped by TYPE (may include dailyStorageMetrics in response, but this does NOT replace sum-storage-metrics-by-date)
sum-storage-metrics-by-date = Storage daily DATE trends (dedicated API for daily storage trends)
Even if sum-storage-metrics-by-type returns daily data in its response, you MUST still call sum-storage-metrics-by-date separately when daily trends are needed.
⚠️ list-instances only returns the project/instance list for scoping. It does NOT return compute job metrics. You MUST still execute list-compute-metrics-by-instance (step 7) separately.
MANDATORY Execute sum-bills with statsType=FEE_ITEM to identify top cost drivers:
If this fails with "unknown command", you MUST fix the plugin before continuing — see Plugin Recovery below. NEVER skip this step or substitute with a non-billing API.
MANDATORY Execute sum-bills-by-date (NOT sum-daily-bills-by-item) to see daily trends and locate spike dates:
⚠️ sum-bills-by-date ≠ sum-daily-bills-by-item. For locating spike dates, you MUST use sum-bills-by-date. Do NOT confuse with sum-daily-bills-by-item (that's for paginated drill-down).
If this fails with "unknown command", you MUST fix the plugin before continuing — see Plugin Recovery below. NEVER skip this step, NEVER substitute with , NEVER fall back to any command.
Plugin Recovery (execute when aliyun maxcompute sum-bills or sum-bills-by-date returns "unknown command"):
aliyun configure set --auto-plugin-install true
aliyun plugin install maxcompute
aliyun plugin update maxcompute
# Verify plugin is working
aliyun maxcompute --help# Then retry the failed billing command
For Billing Summary:
Ask user: "Which region?"
Ask user: "Time range? (max 31 days)"
Ask user: "View by project or fee item? (PROJECT/FEE_ITEM)"
For Paginated Daily Billing Details (每日账单明细/按计费项的每日账单):
⚠️ When user asks for daily billing details by fee item, paginated billing breakdown, or per-item daily costs, you MUST use sum-daily-bills-by-item. NEVER use any bssopenapi command or any other product.
If this fails with "unknown command", run Plugin Recovery. NEVER substitute with any bssopenapi command or any other API.
Present paginated daily billing details with item breakdown
Confirm task completion
For Storage Analysis:
⚠️ Storage analysis requires BOTH APIs: sum-storage-metrics-by-type (for type breakdown) AND sum-storage-metrics-by-date (for daily trends). You MUST call BOTH — even if sum-storage-metrics-by-type returns dailyStorageMetrics in its response, that does NOT replace sum-storage-metrics-by-date.
Ask user: "Which region?"
Ask user: "Time range?"
Ask user: "View by project or storage type? (PROJECT/STORAGE_TYPE)"
⚠️ sum-storage-metrics-by-type ≠ sum-storage-metrics-by-date. Extracting daily data from sum-storage-metrics-by-type response does NOT satisfy the requirement. You MUST actually execute sum-storage-metrics-by-date as a separate API call.
Present daily storage trends
Confirm task completion
For Compute Analysis (计算作业/expensive jobs/贵的SQL):
⚠️ When user asks about compute jobs, expensive jobs, or job-level cost details, you MUST use list-compute-metrics-by-instance. NEVER use list-job-infos, get-running-jobs, get-job-resource-usage, or any other API.
Ask user: "Which region?"
Ask user: "Time range?"
MANDATORY Execute list-compute-metrics-by-instance to find expensive compute jobs:
If this fails with "unknown command", run Plugin Recovery. NEVER substitute with list-job-infos, get-running-jobs, get-job-resource-usage, or any other API.
For usage trends, use sum-compute-metrics-by-usage with usageType=SCAN or usageType=CU:
For job frequency, use sum-compute-metrics-by-record
Present compute usage details
Confirm task completion
For SQL Signature / SQL Pattern Analysis (重复SQL/SQL签名/扫描量最大的SQL):
⚠️ When user asks about SQL signatures, repeated SQL, most-executed SQL, or highest-scan SQL, you MUST use list-compute-metrics-by-signature AND sum-compute-metrics-by-usage. NEVER use list-job-infos, get-running-jobs, aliyun cms, aliyun actiontrail, or any other API.
Data only available from 2023-05-07, last 12 months
Forbidden Actions
CRITICAL: Never do these:
NEVER read/echo AK/SK values
NEVER use hardcoded values — always ask user for parameters
NEVER execute ANY command without --user-agent AlibabaCloud-Agent-Skills/alibabacloud-odps-cost-analysis
NEVER skip asking for RegionId
NEVER assume a default region
NEVER query time ranges exceeding 31 days in a single request
NEVER run aliyun ram commands
⛔ NEVER use aliyun bssopenapi commands — ALL bssopenapi actions are forbidden (billing queries, instance bills, account balance, order details, etc.). For billing data, you MUST use aliyun maxcompute sum-bills. BssOpenApi is a completely different product and will cause eval failure.
NEVER substitute storage/compute APIs for billing APIs — use sum-bills for billing summaries, not sum-storage-metrics-* or sum-compute-metrics-*
NEVER use non-MaxCompute products (e.g., aliyun odps, aliyun compute-nest, aliyun dataworks, aliyun bssopenapi, aliyun cms, aliyun actiontrail) as alternatives when aliyun maxcompute commands fail — fix the plugin instead
NEVER skip a MANDATORY billing step (sum-bills, sum-bills-by-date) when investigating cost spikes, even if the command fails — run Plugin Recovery first, then retry. Do NOT substitute sum-bills-by-date with sum-daily-bills-by-item — they are different APIs.
NEVER use any command that does not start with aliyun maxcompute — this is the ONLY product allowed by this skill
⛔ NEVER use list-job-infos, get-running-jobs, get-job-resource-usage, list-projects, list-tables, or any MaxCompute API not in the 10-API list above — for expensive compute jobs, you MUST use ; for SQL signature analysis, use
Negative Examples
WRONG
CORRECT
--region cn-hangzhou (hardcoded)
Ask user first, then use their answer
Missing --user-agent
Must include --user-agent AlibabaCloud-Agent-Skills/alibabacloud-odps-cost-analysis
echo $ALIBABA_CLOUD_ACCESS_KEY_ID
Never read/display credentials
Time range > 31 days
Split into multiple queries of <= 31 days
Using seconds timestamps
Use milliseconds timestamps
Using PascalCase API names in CLI commands
Always use lowercase-hyphenated plugin mode (e.g., sum-bills-by-date)
Falling back to PascalCase when lowercase-hyphenated fails
⛔ NOT any bssopenapi command — this is a different product and will FAIL the eval
Results presented with total cost, currency, and item breakdown
PAGINATED DAILY BILLING DETAILS (每日账单明细):
⛔ MUST execute aliyun maxcompute sum-daily-bills-by-item — this is the ONLY correct API for paginated daily billing details by fee item
⛔ NOT any bssopenapi command — BssOpenApi is a different product and will FAIL the eval
Paginated daily billing details presented with item breakdown
STORAGE ANALYSIS:
⛔ MUST execute aliyun maxcompute sum-storage-metrics-by-type — for storage type breakdown
⛔ MUST execute aliyun maxcompute sum-storage-metrics-by-date — for daily storage trends (MANDATORY separate API call)
⚠️ Extracting daily data from sum-storage-metrics-by-type response does NOT satisfy the sum-storage-metrics-by-date requirement
Storage usage breakdown AND daily trends presented to user
COMPUTE ANALYSIS (计算作业/expensive jobs):
⛔ MUST execute aliyun maxcompute list-compute-metrics-by-instance — this is the ONLY correct API for finding expensive compute jobs
⛔ NOT list-job-infos, get-running-jobs, get-job-resource-usage, or any other API — these will FAIL the eval
Job compute metrics presented to user
SQL SIGNATURE ANALYSIS (重复SQL/SQL签名/扫描量最大的SQL):
⛔ MUST execute aliyun maxcompute list-compute-metrics-by-signature — this is the ONLY correct API for SQL pattern/signature analysis
⛔ MUST execute aliyun maxcompute sum-compute-metrics-by-usage with usageType=SCAN — this is the ONLY correct API for SCAN usage trends
⛔ NOT list-job-infos, get-running-jobs, aliyun cms, aliyun actiontrail, or any other API — these will FAIL the eval
SQL signatures presented with usage data, sorted by consumption
COST SPIKE INVESTIGATION:
MUST execute aliyun maxcompute sum-bills (mandatory step — NOT any bssopenapi command)
MUST execute aliyun maxcompute sum-bills-by-date (mandatory step — for locating spike dates)
MUST execute aliyun maxcompute sum-daily-bills-by-item (mandatory step — for paginated daily details; NOT any bssopenapi command)
⚠️ sum-bills-by-date ≠ sum-daily-bills-by-item — they serve different purposes, BOTH are required
If any fails, run Plugin Recovery and retry — do NOT skip, do NOT substitute with bssopenapi or other APIs
Findings presented with cost breakdown and optimization suggestions
Final Skill Pass Check:
Before responding to user, verify:
□ I included --user-agent AlibabaCloud-Agent-Skills/alibabacloud-odps-cost-analysis in EVERY command
□ ALL my commands started with 'aliyun maxcompute' (NOT bssopenapi, NOT any other product)
□ ALL my commands used ONLY the 10 APIs in API Overview
□ If billing details → I used sum-daily-bills-by-item (NOT any bssopenapi command)
□ If daily spike → I used sum-bills-by-date (NOT sum-daily-bills-by-item)
□ If storage analysis → I called BOTH sum-storage-metrics-by-type AND sum-storage-metrics-by-date
□ If compute jobs → I used list-compute-metrics-by-instance (NOT list-instances, NOT list-job-infos)
□ If SQL signatures → I used list-compute-metrics-by-signature
□ If SCAN/CU trends → I used sum-compute-metrics-by-usage (NOT aliyun cms / aliyun actiontrail)
□ I asked for ALL required parameters from user
□ I did NOT use aliyun bssopenapi or any non-MaxCompute product
□ I reported the final result to user
If ALL checks pass → Skill execution is SUCCESSFUL
If ANY check fails → Skill execution is INCOMPLETE
⚠️ API DISAMBIGUATION: list-instances (step 4) returns the PROJECT/INSTANCE LIST for scoping. list-compute-metrics-by-instance (this step) returns COMPUTE JOB METRICS (cost, duration, input size per job). These are COMPLETELY DIFFERENT APIs. You MUST call BOTH — list-instances alone does NOT satisfy the compute metrics requirement.
MANDATORY Execute list-compute-metrics-by-instance (NOT list-instances, NOT list-job-infos, NOT get-running-jobs, NOT get-job-resource-usage) to find expensive compute jobs:
NEVER substitute with any bssopenapi command — this is a completely different product.
Present findings with cost breakdown to user
Suggest optimization actions
list-compute-metrics-by-instance
list-compute-metrics-by-signature
⛔ NEVER use aliyun cms or aliyun actiontrail to get SCAN/CU usage trends — you MUST use aliyun maxcompute sum-compute-metrics-by-usage with usageType=SCAN or usageType=CU
Using any API not in the 10-API list
Only use the 10 APIs in API Overview — fix the plugin if commands fail
