Index of every page on this site: llms.txt
All skills
andrelandgraf avatar

/workflow-best-practices

@4449331
Run it yourself

Build durable workflows with the Workflow Development Kit — steps, streaming, agent runs, metadata, and persistence. Use when authoring, starting, resuming, or persisting a workflow.

Use this Skill: https://skilld.dev/gh/andrelandgraf/fullstackrecipes/workflow-best-practices

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈52 tokens for metadata: the name and description. ≈1k when used: this file.

Description uses 2.6% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈208 characters, or 2.6%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

Workflow Best Practices

Build durable workflows with steps, streaming, and agent execution.

Prerequisites

Complete these setup recipes first:

  • Workflow Development Kit Setup

Folder Structure

Each workflow gets a subfolder under src/workflows/. index.ts orchestrates ("use workflow"); steps/ holds durable checkpoints ("use step"); shared steps live in the top-level steps/.

src/workflows/
  steps/           # shared step functions (e.g. stream helpers)
    stream.ts
  chat/
    index.ts       # orchestration ("use workflow")
    steps/         # workflow-specific steps ("use step")
      history.ts
      logger.ts
      name-chat.ts
    types.ts       # UI message types

Creating a Workflow

The orchestration function carries "use workflow" and calls steps. Always wrap an agent run with startStream(messageId) before and finishStream() after — WorkflowChatTransport needs the start/finish frames to parse the response.

// src/workflows/chat/index.ts
import { getWorkflowMetadata, getWritable } from "workflow";
import { startStream, finishStream } from "../steps/stream";
import { chatAgent } from "@/lib/ai/chat-agent";

export async function chatWorkflow({ chatId, userMessage }) {
  "use workflow";

  const { workflowRunId } = getWorkflowMetadata();

  await persistUserMessage({ chatId, message: userMessage });

  // runId lets clients resume this stream later
  const messageId = await createAssistantMessage({
    chatId,
    runId: workflowRunId,
  });
  const history = await getMessageHistory(chatId);

  await startStream(messageId);
  const { parts } = await chatAgent.run(history, {
    maxSteps: 10,
    writable: getWritable(),
  });
  await persistMessageParts({ chatId, messageId, parts });
  await finishStream();

  await removeRunId(messageId);
}

startStream writes { type: "start", messageId }; finishStream writes { type: "finish", finishReason: "stop" } and closes the writable.

Steps

Steps are durable checkpoints that persist their results.

async function getMessageHistory(chatId: string) {
  "use step";

  const dbMessages = await getChatMessages(chatId);
  return convertDbMessagesToUIMessages(dbMessages);
}

The workflow runtime can't import Node modules, so wrap logger calls in a step.

// src/workflows/chat/steps/logger.ts
import { logger } from "@/lib/logging/logger";

export async function log(
  level: "info" | "warn" | "error" | "debug",
  message: string,
  data?: Record<string, unknown>,
): Promise<void> {
  "use step";

  if (data) {
    logger[level](data, message);
  } else {
    logger[level](message);
  }
}

Starting and Resuming

Start with start from workflow/api; reconnect to an in-progress or completed run with getRun.

import { start, getRun } from "workflow/api";
import { chatWorkflow } from "@/workflows/chat";

const run = await start(chatWorkflow, [{ chatId, userMessage }]);
// run.runId       - unique id for this run
// run.readable    - stream of UI message chunks

const resumed = await getRun(runId);
const readable = await resumed.getReadable({ startIndex });

Persisting Results

Save agent output in a step. assertChatAgentParts narrows the generic UIMessage["parts"] to the app's tool/data types before insert.

// src/workflows/chat/steps/history.ts
import type { UIMessage } from "ai";
import { insertMessageParts } from "@/lib/chat/queries";
import { assertChatAgentParts } from "../types";

export async function persistMessageParts({
  chatId,
  messageId,
  parts,
}: {
  chatId: string;
  messageId: string;
  parts: UIMessage["parts"];
}): Promise<void> {
  "use step";

  assertChatAgentParts(parts);
  await insertMessageParts(chatId, messageId, parts);

  await db
    .update(chats)
    .set({ updatedAt: new Date() })
    .where(eq(chats.id, chatId));
}

References

Source: SKILL.md on GitHub

No rule matched.

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No alerts5mo3 checks · Risk SAFE
  • Gen Agent Trust Hub5mo

    This skill provides a comprehensive guide and code examples for building durable workflows using the Workflow Development Kit. It covers architectural patterns, streaming, and persistence without introducing security risks.

  • Socket5mo

    No alerts

  • Snyk5mo

    Risk: LOW · No issues

Signed by skilld at 4449331. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 3 weeks ago.

Steadyupdated 6 months ago

README badge

README badge for andrelandgraf/fullstackrecipes/workflow-best-practices