agentleFS
kumaran-iskumaran-is/claude-code-onboarding/.claude/skills/workflow-orchestration-patterns/SKILL.md

workflow-orchestration-patterns

Skill35 starsChanged 6 months agoMITView on GitHub

Durable workflow orchestration with Temporal for distributed systems across Java 21, Python 3.14, and NestJS 11.x. Covers Workflow vs Activity separation, Saga pattern, Entity workflows, Fan-out/Fan-in, determinism constraints, retry policies, and idempotency. Use when building long-running, failure-resilient distributed business processes.

  • Installs packages

7 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 7 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Before generating code, consult these sources for current SDK APIs:

    line 86
  2. Distinctive:

    Idempotency rule: every activity MUST be safe to call N times.

    line 132
  3. Distinctive:

    After writing Temporal workflow code, dispatch these reviewer agents:

    line 136

“Common” is measured: kept by at least 6 in 10 of the 7 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Iron Law

NO WORKFLOW WITHOUT DETERMINISM ENFORCEMENT AND IDEMPOTENT ACTIVITIES — non-deterministic workflow code causes replay failures; non-idempotent activities cause duplicate charges, double-sends, and data corruption

Workflow Orchestration Patterns — Temporal + Java 21 / Python 3.14 / NestJS 11.x

Quick Scaffold

# Start Temporal server locally (Docker)
docker run -d --name temporal \
  -p 7233:7233 -p 8080:8080 \
  temporalio/auto-setup:1.24

# Java 21 / Spring Boot — add to pom.xml
# io.temporal:temporal-sdk:1.25.0
# io.temporal:temporal-spring-boot-autoconfigure-alpha:0.6.0

# Python 3.14 — uv add
uv add temporalio==1.7.0 fastapi uvicorn

# NestJS 11.x — npm
npm install @temporalio/client @temporalio/worker @temporalio/workflow @temporalio/activity

Process

  1. Deploy Temporal — local Docker or Temporal Cloud (cloud.temporal.io)
  2. Define Activities — external interactions (API calls, DB writes, emails) — must be idempotent
  3. Define Workflow — orchestration logic only — must be deterministic (no I/O, no Date.now(), no random)
  4. Configure Retry Policies — initial interval, backoff coefficient, max attempts, non-retryable errors
  5. Register Worker — binds workflow + activity implementations to task queues
  6. Start Execution — client sends startWorkflow() with typed input
  7. Add Signals/Queries — external state mutations (signals) and reads (queries) on running workflows
  8. Implement Saga — register compensations BEFORE each step; run in reverse (LIFO) on failure
  9. Add Heartbeats — long-running activities must call heartbeat() periodically
  10. Write Tests — use Temporal test environment with time-skipping for deterministic test execution

Key Patterns

PatternWhen to UseReference
Saga + CompensationDistributed transactions needing rollbackreference/implementation-playbook.md#saga
Entity Workflow (Actor)One workflow per entity lifecycle (cart, account, order)reference/implementation-playbook.md#entity
Fan-Out / Fan-InParallel execution of N tasks with result aggregationreference/implementation-playbook.md#fanout
Async Callback / SignalWaiting for external event or human approvalreference/implementation-playbook.md#signals
Activity HeartbeatLong-running activities (>30s) with progress trackingreference/implementation-playbook.md#heartbeat
Workflow VersioningSafe code changes while old executions still runningreference/implementation-playbook.md#versioning
Child WorkflowsDecompose large workflows for scalability (1M tasks = 1K x 1K)reference/implementation-playbook.md#child

Workflow vs Activity Decision Rule

Does the code touch an external system? (API, DB, file, email, network)
  YES -> Activity
  NO  -> Workflow (orchestration/decision logic only)

Prohibited in Workflow code:

  • new Date() / datetime.now() / LocalDateTime.now() — use Workflow.currentTimeMillis() / workflow.now()
  • Math.random() / random.random() — use Workflow.newRandom() / workflow.random()
  • HTTP calls, DB queries, file I/O — move to Activity
  • Threading, locks, Thread.sleep() — use Workflow.sleep()
  • Non-deterministic libraries

Conventions & Rules

For per-SDK implementation templates (Java, Python, NestJS), read reference/implementation-playbook.md

Documentation Sources

Before generating code, consult these sources for current SDK APIs:

SourceURL / ToolPurpose
Temporal Docshttps://docs.temporal.io/dev-guideCore concepts, SDK APIs, best practices
Java SDKhttps://www.javadoc.io/doc/io.temporal/temporal-sdk/latest/Java workflow/activity annotations
Python SDKhttps://python.temporal.io/Python asyncio API reference
TypeScript SDKhttps://typescript.temporal.io/NestJS/TypeScript worker and client APIs
Context7 MCPresolve-library-id: temporalioLatest Temporal Python patterns

Reference Files

FileContentWhen to Use
reference/implementation-playbook.mdFull workflow + activity code for Java, Python, NestJS; Saga, Entity, Fan-out, Signal, Heartbeat, Versioning, Child Workflow patternsAny Temporal implementation

Common Commands

# View Temporal Web UI (workflows, history, task queues)
open http://localhost:8080

# Java — run worker
mvn spring-boot:run

# Python — run worker
uvicorn src.main:app --reload &
python -m src.worker   # separate process

# NestJS — run worker
npm run start:worker   # defined in package.json scripts

# Run Temporal test suite (time-skipping enabled)
# Java: mvn test
# Python: pytest -q
# NestJS: npm test

Error Handling

For retry policy templates and non-retryable error classification per SDK, read reference/implementation-playbook.md#retry

Activity retry rule: classify every exception before throwing. Validation errors and business rule violations -> ApplicationFailure.newNonRetryableFailure(). Transient network/timeout errors -> retryable (default).

Workflow failure rule: workflows do not catch activity exceptions unless implementing compensation. Let Temporal's retry engine handle transient failures. Only catch ActivityFailure when running saga compensations.

Idempotency rule: every activity MUST be safe to call N times. Use idempotency keys, upsert patterns, and check-then-act with unique constraints. Verify with: "If this activity runs twice with the same input, is the final state identical?"

Post-Code Review

After writing Temporal workflow code, dispatch these reviewer agents:

  • architect-review — workflow boundaries, saga completeness, entity lifecycle correctness
  • security-reviewer — activity input validation, no secrets in workflow state, payload size limits (2MB)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.