conversion-task-plan-rules

name	conversion-task-plan-rules
description	Rules for generating ordered conversion task plans. Covers mandatory task ordering, required task types, descriptive task ID rules, optional vs non-optional tasks, output path conventions, and what each task must contain.

Skill: Conversion Task Plan Rules

Purpose: Authoritative rules for how an AI agent should generate the ordered task plan for converting a flow. Follow exactly.

⚠️ DECOMPILATION PREREQUISITE: Whenever source behavior required to fully implement the workflows exists only in compiled assemblies (.dll/.exe), you MUST follow skill dependency-and-decompilation-analysis to decompile BEFORE generating conversion tasks. This applies to orchestrations and any workflow-critical logic — custom pipeline components, helper libraries, functoids, map extension code, and shared business rules. Do NOT generate tasks until all relevant assemblies are decompiled and their logic is understood.

---

1. Task Derivation

After reading planning results (migration_conversion_getPlanningResults), determine ALL conversion tasks needed based on the specific flow's planned architecture.

---

2. Mandatory Task Order

2.1 Task 1 — Scaffold (REQUIRED)

• Type: scaffold-project
• MUST be the first task.
• Read the skill scaffold-logic-apps-project and follow it exactly.
• Do NOT inline scaffold details — just reference the skill.

2.2 Artifact Tasks (Middle)

After scaffold, add tasks for:

• Schemas → {outputLogicAppRoot}/Artifacts/Schemas/
• Maps → {outputLogicAppRoot}/Artifacts/Maps/
• Rules → {outputLogicAppRoot}/Artifacts/Rules/
• Certificates → {outputLogicAppRoot}/Artifacts/Certificates/
• Local functions → using skill dotnet-local-functions-logic-apps
• Workflows → {outputLogicAppRoot}/<Name>/workflow.json
• Connections → {outputLogicAppRoot}/connections.json
• Cloud resource provisioning (only for connectors with no local alternative)

IMPORTANT Integration Account decision rule:

• For every artifact-related task, the agent MUST explicitly think through whether an Integration Account is required.
• Use an Integration Account when the design requires Integration Account-only capabilities, especially B2B/EDI scenarios such as X12, EDIFACT, AS2, trading partners, agreements, or centralized shared artifact management.
• If an Integration Account is used for the flow, then ALL deployable artifacts for that flow must be planned for the Integration Account path/use model consistently.
• Do NOT split artifacts between Integration Account and the Logic App artifact folders. These two models are mutually exclusive for a given flow plan.
• If an Integration Account is used, create/provision/configure the Integration Account BEFORE any schema/map/certificate/trading-partner artifact task that depends on it.
• If an Integration Account is used, the Integration Account provisioning task MUST actually deploy/create the Integration Account resource in Azure in that task itself. It must NOT stop at generating deployment scripts or templates only.
• That Integration Account provisioning task MUST retrieve the deployed Integration Account resource ID and callback URL, then update local.settings.json with WORKFLOWS_INTEGRATION_ACCOUNT_ID and WORKFLOW_INTEGRATION_ACCOUNT_CALLBACK_URL in that same task.
• The NEXT Integration Account-related artifact task MUST upload the required schemas/maps/certificates/partners/agreements into the Integration Account itself. Do NOT leave artifact upload implicit or deferred.
• CRITICAL — Agreement Schema References: When EDIFACT/X12 agreements are created, their schemaReferences arrays MUST NOT be left empty. After schemas are uploaded to the Integration Account, the agreements MUST be updated via PATCH or re-PUT so schemaReferences links each agreement to the correct message schemas for decode/encode. Empty schemaReferences causes EdifactDecode or X12Decode actions to fail at runtime with UnexpectedSegment errors. The schema-upload task's executionPrompt MUST explicitly instruct the agent to update agreement schemaReferences after uploading schemas.
• If the flow does NOT require an Integration Account, then use the Logic App artifact folders only.
• If an Integration Account is used, the plan MUST also account for the required Logic Apps Standard app settings: WORKFLOWS_INTEGRATION_ACCOUNT_ID, with the Integration Account resource ID as its value (example: /subscriptions/{subId}/resourceGroups/{rg}/providers/Microsoft.Logic/integrationAccounts/{name}), and WORKFLOW_INTEGRATION_ACCOUNT_CALLBACK_URL, with the correct Integration Account callback URL value.
• Any Integration Account provisioning/configuration task MUST put that app setting requirement directly into its executionPrompt, not only in the high-level summary.
• The task plan MUST make this choice explicit in the artifact task descriptions or execution prompts.

2.3 Runtime Validation (REQUIRED)

• Type: validate-runtime
• Must come after all generation tasks.
• Runs func start --verbose, checks for errors, fixes issues, re-runs until clean.

2.4 Local E2E Testing (REQUIRED)

• Type: test-workflows
• Name: "End to End Testing (Local Environment)"
• Must come after runtime validation.
• Full test matrix (see skill runtime-validation-and-testing).

2.5 Local Black Box Testing (OPTIONAL)

• Type: local-blackbox-test
• Name: "Local Black Box Testing (Only if test data provided)"
• Must come after local E2E testing (test-workflows).
• Mark as optional: true — user decides whether to execute or skip.
• When the user executes this task, the extension prompts for a test data folder. That folder MUST contain:
  • Input/output content files representing real or expected integration payloads.
  • A TEST-INSTRUCTIONS.md file with source-platform-specific instructions for how the agent should validate the Logic App against those input/output files.
• Validation: The extension will reject the selected folder and abort the task if TEST-INSTRUCTIONS.md is not found in it. The user must re-select a valid folder.
• The agent MUST:
  1. Read TEST-INSTRUCTIONS.md from the provided folder.
  2. Understand the instructions — they may reference the source platform (BizTalk, MuleSoft, etc.) and describe expected transformations, routing, enrichments, or validations.
  3. For each input file, send it to the appropriate Logic App workflow trigger (HTTP, file, queue, etc.).
  4. Compare the actual output against the expected output files.
  5. Report pass/fail for each test case with details on mismatches.
  6. MANDATORY: Generate BLACKBOX-TEST-REPORT.md in the project root with: test date, flow name, per-test-case results (input file, expected output, actual result, pass/fail, mismatch details), and a summary table with total/passed/failed counts. Do NOT consider the task complete until this file is created.
• The executionPrompt for this task MUST include: "Read the TEST-INSTRUCTIONS.md from the user-provided test data folder first, then execute all tests described in it against the local Logic App runtime. After all tests, generate BLACKBOX-TEST-REPORT.md in the project root."

2.6 Cloud Testing (OPTIONAL — LAST)

• Type: cloud-deploy-test
• Name: "End to End Testing (Cloud Environment) (Only if required)"
• MUST be the ABSOLUTE LAST task.
• Mark as optional: true — user decides whether to execute or skip.
• If a BLACKBOX-TEST-REPORT.md exists in the project root (generated by the local black box testing task), the cloud testing agent MUST also read it and re-execute all test cases from that report against the deployed Azure Logic App, in addition to the standard TEST-REPORT.md scenarios.

---

3. Task ID Rules

• Task IDs MUST be descriptive: e.g. scaffold-project, convert-schemas, generate-workflow-receive-order, provision-servicebus, validate-runtime, test-workflows, local-blackbox-test, cloud-deploy-test.
• NEVER use generic sequential IDs like task-1, task-2 — they will be REJECTED by the tool.

---

4. Required Task Fields

Each task needs:

Field	Type	Required	Description
id	string	Yes	Descriptive unique ID
type	string	Yes	Short type label
name	string	Yes	Human-readable name
description	string	Yes	Concise user-facing summary (1-2 lines)
executionPrompt	string	No	Detailed agent-only instructions for execution
dependsOn	string[]	Yes	Prerequisite task IDs
order	number	Yes	1-based execution sequence
artifactIds	string[]	No	Artifact IDs this task operates on
estimatedMinutes	number	No	Effort estimate
optional	boolean	No	True for user-skippable tasks (e.g. cloud-deploy)

---

5. Output Path Conventions

• Workspace root: {outputProjectRoot}/
• Logic App project: {outputLogicAppRoot}/
• Workspace file: {outputProjectRoot}/{outputProjectName}.code-workspace
• Every task description MUST state the exact output file path relative to the project root.
• NEVER scaffold into the source project root.

---

6. Key Rules for Task Descriptions

• For each task, set description to a concise user-facing summary (1-2 lines).
• Put all detailed implementation steps/rules into executionPrompt.
• Workflow generation tasks MUST include: "Read skill source-to-logic-apps-mapping first, then search reference workflows."
• Connection generation tasks MUST include: "Search with category=connection to find exact connections.json format."
• Local function tasks MUST reference skill dotnet-local-functions-logic-apps.
• ALL code generation tasks MUST follow skill no-stubs-code-generation.