1. Diagnose PAI-Rec Engine Interface Issues: When engine API returns errors or unexpected results, trace the request through EAS service logs and engine configurations to identify root causes.
2. Validate Engine Configurations: Analyze engine configuration files for potential issues, inconsistencies, or misconfigurations before deployment.

• PAI-EAS Service: Elastic Algorithm Service hosting the recommendation engine
• PAI-Rec Engine: The recommendation engine processing requests
• Engine Configuration: Configuration files defining engine behavior
• Service Logs: EAS service logs containing request traces

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 install/update, or see

references/cli-installation-guide.md

for installation instructions.

[MUST] run

aliyun configure set --auto-plugin-install true

to enable automatic plugin installation. [MUST] run

aliyun plugin update

to ensure that any existing plugins are always up-to-date.

Pre-check: Alibaba Cloud Credentials Required

Security Rules:

• NEVER read, echo, or print AK/SK values (e.g.,

  echo $ALIBABA_CLOUD_ACCESS_KEY_ID

  is FORBIDDEN)
• NEVER ask the user to input AK/SK directly in the conversation or command line
• NEVER use

  aliyun configure set

  with literal credential values
• ONLY use

  aliyun configure list

  to check credential status

bash

aliyun configure list

Check the output for a valid profile (AK, STS, or OAuth identity).

If no valid profile exists, STOP here.

1. Obtain credentials from Alibaba Cloud Console
2. Configure credentials outside of this session (via

   aliyun configure

   in terminal or environment variables in shell profile)
3. Return and re-run after

   aliyun configure list

   shows a valid profile

[MUST] Permission Failure Handling: When any command or API call fails due to permission errors at any point during execution, follow this process:

1. Read

   references/ram-policies.md

   to get the full list of permissions required by this SKILL
2. Use

   ram-permission-diagnose

   skill to guide the user through requesting the necessary permissions
3. Pause and wait until the user confirms that the required permissions have been granted

IMPORTANT: Parameter Confirmation — Before executing any command or API call, ALL user-customizable parameters (e.g., RegionId, service names, instance IDs, environment names, config names, etc.) MUST be confirmed with the user. Do NOT assume or use default values without explicit user approval.

At the start of the Core Workflow (before any CLI invocation): [MUST] Enable AI-Mode — AI-mode is required for Agent Skill execution. Run the following commands before any CLI invocation:

bash

aliyun configure ai-mode enable
aliyun configure ai-mode set-user-agent --user-agent "AlibabaCloud-Agent-Skills/alibabacloud-pai-rec-diagnosis"

[MUST] Disable AI-Mode at EVERY exit point — Before delivering the final response for ANY reason, always disable AI-mode first. This applies to ALL exit paths: workflow success, workflow failure, error/exception, user cancellation, session end, or any other scenario where no further CLI commands will be executed. AI-mode is only used for Agent Skill invocation scenarios and MUST NOT remain enabled after the skill stops running.

bash

aliyun configure ai-mode disable

aliyun eas describe-service \
  --cluster-id <cluster-id> \
  --service-name <service-name>

• Resource

  : EAS service resource ID (e.g.,

  eas-r-1v4qb1yan3qmnjwxqe

  )

• ServiceConfig.envs

  : Environment variables containing:

  • REGION

    : The region

  • INSTANCE_ID

    : PAI-Rec instance ID

  • CONFIG_NAME

    : Engine configuration name

  • PAIREC_ENVIRONMENT

    : Environment (product/prepub)

--start-time

--end-time

aliyun eas describe-service-log \
  --cluster-id <cluster-id> \
  --service-name <service-name> \
  --keyword <request-id> \
  --page-size 500

• When only

  --keyword

  is supplied (no time range), the CLI returns the full PAI-Rec application trace (

  controller.go

  /

  feed.go

  /

  recall.go

  /

  rank_service.go

  etc.) matching the request_id.
• As soon as

  --start-time

  /

  --end-time

  are added — even if the window covers the real log timestamp — the CLI silently drops business logs and only returns infrastructure noise (

  /bin/sh

  wrapper heartbeats,

  502 Bad Gateway

  retries,

  postgres.go dbstat

  ).
• Therefore: for request-level diagnosis, always omit the time range and rely on

  --keyword <request-id>

  alone.

• --keyword

  : Use the full

  request_id

  extracted from the API response (case-sensitive exact match).

• --page-size

  : Raise to 500 to capture the entire trace in a single page; total matched entries for one request is usually < 30.

• --start-time

  /

  --end-time

  : Only use these for broad time-window scans without

  --keyword

  (e.g., when investigating non-request-specific issues). Required format is

  yyyy-MM-dd HH:mm:ss

  in UTC (space separator, no

  T

  / no

  Z

  ). ISO-8601 forms like

  2025-04-28T00:00:00Z

  will be rejected with

  InvalidParameter

  .

• product

  →

  Prod

• prepub

  →

  Pre

aliyun pairecservice list-engine-configs \
  --instance-id <instance-id> \
  --environment <Prod|Pre> \
  --status Released \
  --name <config-name>

• Find the configuration with

  Status: Released

• Get

  EngineConfigId

  and

  Version

aliyun pairecservice get-engine-config \
  --instance-id <instance-id> \
  --engine-config-id <engine-config-id>

• ConfigValue

  : The actual engine configuration (JSON/YAML)

scripts/validate.py

ConfigValue

printf '%s' "$CONFIG_VALUE" | python3 scripts/validate.py --stdin

RecallConfs

FilterConfs

SceneConfs

scene_id

SceneConfs

validate.py

• validate.py

  findings may enter the final diagnosis ONLY when they are directly tied to the log evidence for the current

  request_id

  (e.g. the log blames a

  RecallConf

  name that

  validate.py

  flags as duplicated or dangling).
• Findings unrelated to the current

  request_id

  trace MUST NOT be added to the final conclusion. They remain an internal sanity-check signal only. This preserves the evidence-only reporting rule in Step 6.

1. API Response: Error code, message, and returned data
2. Service Logs: Trace logs for the request_id showing processing flow
3. Engine Configuration: Settings that may affect the behavior

• Configuration mismatches (e.g., recall settings, filtering rules)
• Resource limitations (e.g., insufficient items, timeout settings)
• Data source issues (e.g., table access, feature availability)
• Environment inconsistencies (e.g., prod config in prepub environment)

• Report only what is observed. Quote the exact log line (file:line, level, message) and the exact config fragment that proves each claim.
• State the direct causal chain from log evidence to the API response, and stop there.
• Do NOT add any of the following unless the user explicitly asks:
  • Speculative root causes not visible in logs/config (e.g., "client probably sent wrong X")
  • Fix recommendations or remediation steps
  • Conditional "if X then Y" scenarios
  • Tangential best-practice advice (security, fallback design, naming, etc.)
  • Guesses about upstream systems, client code, or data sources not covered by the logs/config
• If the evidence is insufficient to reach a conclusion, state explicitly what additional data (specific log lines, other config versions, other environments) is needed, instead of guessing.
• Recommendations are opt-in only. Provide fixes/suggestions only when the user explicitly requests them in a follow-up.

engine-config-id

aliyun pairecservice list-engine-configs \
  --instance-id <instance-id> \
  --environment <Prod|Pre> \
  --name <config-name>

• Version

  : Version number

• Status

  : Configuration status (Released/Draft/Archived)

• GmtCreateTime

  : Creation timestamp

• EngineConfigId

  : Version ID

engine-config-id

ConfigValue

• ✅ Checks passed (Schema clean, references resolved, no rule violations)
• ⚠️ Warnings reported by the script (severity=warning) or inconsistencies observed in

  ConfigValue

  (e.g. naming collisions between

  RankScore

  variables and model output fields, env/region mismatches)
• ❌ Errors reported by the script (severity=error) or missing required fields
• Missing-evidence notes — what extra data (other config versions, model signatures, etc.) would be needed to turn a warning into a confirmed error

1. For Diagnosis Workflow:
   • Service information retrieved successfully
   • Logs found containing the request_id
   • Configuration loaded correctly
   • Root cause identified
2. For Validation Workflow:
   • Configuration retrieved successfully
   • All validation checks executed
   • Issues clearly reported
   • Recommendations provided (if applicable)

1. Always capture request_id: When reporting API issues, include the full response with request_id for accurate log correlation.
2. Log queries — keyword only, no time range: For request-level diagnosis, pass

   --keyword <request_id>

   to

   aliyun eas describe-service-log

   and leave

   --start-time

   /

   --end-time

   unset. Combining keyword with a time range filters out business logs due to a CLI quirk (see Workflow 1, Step 3). Only use time ranges for broad non-request scans, and only with the

   yyyy-MM-dd HH:mm:ss

   UTC format (no

   T

   / no

   Z

   ).
3. Environment awareness: Always verify that configurations match the target environment (Prod vs Pre).
4. Version control: When validating configurations, check multiple versions if issues persist across deployments.
5. Log retention: EAS service logs are retained for limited periods; diagnose issues promptly after occurrence.
6. Configuration backup: Before applying changes based on validation results, ensure current configurations are backed up.
7. Cross-reference: Compare working configurations with problematic ones to identify differences.
8. Service status: Check EAS service status before diagnosing; service-level issues may mask configuration problems.
9. Evidence-only conclusions: Ground every statement in the diagnosis on a specific log line or config fragment. Do not speculate, do not propose fixes, and do not volunteer best-practice advice unless the user explicitly asks. If the evidence is insufficient, say what is missing rather than inferring.
10. Structured analysis: Follow the systematic workflow rather than jumping to conclusions based on error messages alone.
11. Document findings: Keep track of recurring issues and their resolutions for faster future diagnosis.