restate-workflows
$
npx mdskill add joelhooks/joelclaw/restate-workflowsProvides a compatibility alias for substrate-specific ADR-0217 runtime bridge tasks, focusing on external-repo bridging and low-level mechanics.
- Helps with external repository submission, queue-family contracts, and low-level runtime submission mechanics.
- Integrates with Redis for queue management, Dkron for scheduling, and Restate for durable execution.
- Recommends using workflow-rig for new work and restricts usage to verified runtime bridge problems.
- Delivers results through the joelclaw CLI without direct Redis access or internal package imports.
SKILL.md
.github/skills/restate-workflowsView on GitHub ↗
---
name: restate-workflows
displayName: Restate Workflows
description: "Compatibility alias for substrate-specific ADR-0217 runtime bridge work. For new work, load `workflow-rig` first; use this skill only when the task is explicitly about external-repo bridging, queue-family contracts, or low-level runtime submission mechanics."
version: 0.2.0
author: Joel Hooks
tags: [restate, workflows, queue, cli, sandbox, adr-0217, integration]
---
# Restate Workflows
This skill is now a **compatibility alias for substrate work**.
For new work, load **`workflow-rig`** first. Use this skill only when the real problem is the runtime bridge itself: external repo submission, queue-family contracts, or low-level ADR-0217 substrate mechanics.
The boundary is the `joelclaw` CLI.
Do not reach into Redis directly. Do not import private joelclaw packages. Do not pretend the other repo knows internal queue schemas unless you have explicitly verified them.
## The runtime shape
ADR-0217 currently reads best as:
- **Redis is the pressure layer** — queue admission, bursts, pause/resume, family control
- **Dkron is the clock** — scheduled starts
- **Restate is the durable executor** — workflow identity, retries, DAG progress
- **Sandboxed runners are the side-effect boundary** — isolated code-changing work
An external repo should integrate at the edge, not by coupling itself to the middle.
That means:
- submit saved workload artifacts through `joelclaw workload run` when the work has already been shaped at the workload layer
- use `joelclaw queue emit` only as the low-level escape hatch for verified raw queue families
- carry enough metadata for idempotency and tracing
- let joelclaw decide how that request reaches Restate or a sandbox runner
## When to use
Use this skill when a user asks for any of these:
- "send work to the queue"
- "have this repo request background work"
- "bridge this repo into joelclaw"
- "submit sandboxed work"
- "emit a workflow request"
- "wrap joelclaw queue emit"
- "run this in Restate from another repo"
## Non-negotiable rules
- **CLI boundary only.** Prefer `joelclaw workload run` for approved workload artifacts; use `joelclaw queue emit` only when you are intentionally working at the raw queue-family layer.
- **Never talk to Redis directly** from the external repo.
- **Never depend on private joelclaw internals** (`@joelclaw/*`, Redis keys, internal TypeScript types, worker-only contracts) unless the repo itself is the joelclaw monorepo.
- **Keep changes inside the calling repo.** Build a wrapper in that repo; do not assume you can edit joelclaw at the same time.
- **Inspect live help first.** The CLI contract wins over memory.
- **Prefer explicit workflow families.** If there is no verified family for the job, stop and document the proposed family/contract instead of guessing silently.
- **Machine-readable output by default.** The wrapper should return stable JSON that callers can parse.
- **Support dry-run.** The repo should be able to show the exact request without enqueuing it.
## First steps
Run these before designing the bridge:
```bash
joelclaw workload run --help
joelclaw queue --help
joelclaw queue emit --help
```
If the target repo is supposed to monitor results too, also inspect:
```bash
joelclaw jobs status --hours 1 --count 10
joelclaw queue observe --hours 1
```
If you are unsure whether a workflow family already exists, search docs or ask instead of inventing one.
## What to build in the external repo
The default deliverable is a thin wrapper command/module/script around `joelclaw workload run` when the repo can produce a saved workload artifact. Fall back to `joelclaw queue emit` only when you have a verified low-level family and there is no workload artifact to bridge.
Good shapes:
- `scripts/request-work.ts`
- `bin/request-background-work`
- `src/lib/joelclaw/requestWorkflow.ts`
- `src/cli/restate-workflow.ts`
Bad shape:
- raw shell one-liners scattered across the repo with no contract doc or tests
## Minimum request payload
Every request payload should include these fields unless the target family has a stricter contract:
```json
{
"idempotencyKey": "repo-name:task-type:base-sha:input-hash",
"requestId": "uuid-or-stable-derived-id",
"repo": {
"path": "/abs/path/or/null",
"url": "git@github.com:owner/repo.git",
"name": "repo-name",
"baseSha": "abc123",
"branch": "main"
},
"task": {
"kind": "sandboxed-change",
"prompt": "what the runtime should do",
"artifacts": ["patch", "test-report", "summary.json"]
},
"caller": {
"source": "external-repo-name",
"requestedBy": "human-or-agent",
"sessionId": "optional-session-id",
"callback": {
"kind": "poll",
"target": "joelclaw jobs status"
}
},
"meta": {
"dryRun": false,
"createdAt": "2026-03-08T05:00:00.000Z"
}
}
```
### Required semantics
- `idempotencyKey` must be stable for the same intended work request
- `baseSha` must reflect the code the sandbox should start from
- `prompt` must say what success looks like
- `artifacts` must list what the caller expects back
- `caller.source` must identify who emitted the request
## Emission pattern
The wrapper should construct JSON, then call the CLI.
Example pattern:
```bash
joelclaw queue emit <event-family> --data '<json>'
```
Example Node/Bun sketch:
```ts
import { spawn } from "node:child_process";
export async function requestWorkflow(
event: string,
payload: unknown,
dryRun = false,
) {
const data = JSON.stringify(payload);
if (dryRun) {
return {
ok: true,
mode: "dry-run",
event,
payload,
};
}
return await new Promise((resolve, reject) => {
const child = spawn("joelclaw", ["queue", "emit", event, "--data", data], {
stdio: ["ignore", "pipe", "pipe"],
});
let stdout = "";
let stderr = "";
child.stdout.on("data", (chunk) => {
stdout += chunk.toString();
});
child.stderr.on("data", (chunk) => {
stderr += chunk.toString();
});
child.on("close", (code) => {
if (code === 0) {
resolve({ ok: true, mode: "live", event, payload, stdout });
return;
}
reject(
new Error(`joelclaw queue emit failed (${code}): ${stderr || stdout}`),
);
});
});
}
```
## Family selection
Do not hijack an unrelated proof family just because it exists.
Rules:
- `content/updated` is a dogfood family, not a generic bucket for outside repos
- if a verified workflow request family already exists, use it
- if no family exists, document the proposed name and payload contract in the repo instead of guessing
A good stop-and-document line is:
> No canonical workflow-request family was verified in the target runtime. This repo now emits a validated payload object locally and documents the proposed `sandbox/work.requested` contract, but does not enqueue live requests until the runtime family is confirmed.
That is honest. Honesty beats fake integration.
## Required wrapper features
### 1. Dry-run mode
The command must be able to print the exact event + payload without sending it.
Dry-run output should be parseable JSON, for example:
```json
{
"ok": true,
"mode": "dry-run",
"event": "sandbox/work.requested",
"payload": { "...": "..." }
}
```
### 2. Clear errors
Handle these explicitly:
- `joelclaw` binary missing
- invalid JSON/payload serialization
- command non-zero exit
- caller omitted `baseSha`, `prompt`, or `idempotencyKey`
### 3. Machine-readable success output
Return stable JSON on success, not prose.
At minimum:
```json
{
"ok": true,
"mode": "live",
"event": "sandbox/work.requested",
"idempotencyKey": "...",
"stdout": "raw joelclaw output"
}
```
### 4. Local tests
Tests should verify:
- payload construction
- idempotency key stability
- dry-run behavior
- CLI invocation shape
- failure handling
Mock the CLI process if needed. Do not require the real runtime for unit tests.
## README contract
Document these four things in the calling repo:
1. what command submits work
2. what event family it emits
3. what payload fields are required
4. how to verify the request from joelclaw surfaces
Minimum verification section:
```bash
joelclaw jobs status --hours 1 --count 10
joelclaw queue observe --hours 1
```
If there is a known workflow ID or downstream surface, include it.
## Anti-patterns
Do not do this:
- writing directly to Redis
- importing joelclaw monorepo code into the external repo
- hiding event-family uncertainty behind vague names like `task/run`
- omitting `baseSha` for code-changing work
- printing unstructured human-only output from the wrapper
- building a wrapper with no dry-run or no tests
- pretending Restate-native execution is already wired if the current path is still transitional
## Recommended deliverables
When asked to implement this integration in another repo, deliver:
1. the wrapper command/module
2. payload contract documentation
3. example invocation
4. verification steps
5. tests
## Short prompt for another agent
Paste this:
> Use `/skill:restate-workflows`. Add a thin wrapper in this repo around `joelclaw queue emit` so it can submit work into joelclaw without talking to Redis or private joelclaw internals. Check `joelclaw queue emit --help` first. Emit machine-readable JSON with `requestId`, `idempotencyKey`, repo `url`/`branch`/`baseSha`, task prompt, expected artifacts, and caller metadata. Support `--dry-run`, add a README example and tests, and if the canonical event family is not verified, stop and document the proposed family instead of guessing.
More from joelhooks/joelclaw
- add-skillCreate new joelclaw skills with the idiomatic process — repo-canonical, symlinked, git-tracked, slogged. Triggers on 'add a skill', 'create skill', 'new skill', 'canonical skill', 'make a skill for', or any request to formalize a process or domain into a reusable skill.
- adr-skillCreate and maintain Architecture Decision Records (ADRs) optimized for agentic coding workflows. Use when you need to propose, write, update, accept/reject, deprecate, or supersede an ADR; bootstrap an adr folder and index; consult existing ADRs before implementing changes; or enforce ADR conventions. This skill uses Socratic questioning to capture intent before drafting, and validates output against an agent-readiness checklist.
- agent-discovery"Optimize websites, docs, and product surfaces for agent discoverability and operator UX. Use when working on agent SEO/AEO/GEO, crawl policy, markdown or JSON projections, llms.txt, sitemap.md, AGENTS.md guidance, content negotiation, accessibility for browser agents, or any request to make a site easier for pi, OpenCode, Claude Code, ChatGPT, Perplexity, or other agent harnesses to find and use."
- agent-loopStart, monitor, and cancel durable multi-agent coding loops via Inngest. Use when the user wants to run autonomous coding workloads, execute a PRD with multiple stories, kick off an AFK coding session, have agents implement features from a plan, or manage running loops. Triggers on "start a coding loop", "run this PRD", "implement these stories", "go AFK and code this", "check loop status", "cancel the loop", "joelclaw loop", or any request for autonomous multi-story code execution.
- agent-mail>-
- agent-workloads"Compatibility alias for the canonical `workflow-rig` front door. Use when older prompts mention `agent-workloads` or when you need the legacy workload-planning guidance; for new work, load `workflow-rig` first."
- clawmail>-
- cli-design"Design and build agent-first CLIs with HATEOAS JSON responses, context-protecting output, and self-documenting command trees. Use when creating new CLI tools, adding commands to existing CLIs (joelclaw, slog), or reviewing CLI design for agent-friendliness. Triggers on 'build a CLI', 'add a command', 'CLI design', 'agent-friendly output', or any task involving command-line tool creation."
- codex-prompting"Use this skill for any request to trigger, coordinate, or craft prompts for Codex. Use when user says 'send to codex', 'use codex', 'prompt codex', 'ask codex', 'delegate to codex', 'run in codex', or asks for a Codex-first execution handoff."
- content-publish"Publish content to joelclaw.com via the Convex-first pipeline. Covers the full lifecycle: draft → review → publish → revalidate → verify. Handles secret leasing, tag conventions, content types (article, tutorial, note, essay), and verification gates. Use when: 'write article about X', 'publish article <slug>', 'draft a tutorial', 'publish this', 'push to convex', or any content publishing task."