---
name: resilience
description: >-
  Apply resilience patterns (retry, backoff, circuit breaking) when writing or reviewing code that makes outbound HTTP/REST calls, database operations, or any I/O that can experience transient failures. Use this skill when: (1) writing new code that calls external services or APIs, (2) reviewing code that uses fetch, HTTP clients, or service-to-service calls, (3) adding queue consumers or workers that interact with remote systems, (4) debugging intermittent failures in production.
metadata:
  version: "1.0.0"
title: resilience
canonical_url: https://skilld.dev/gh/microsoft/scope/resilience
last_updated: 2026-10-19T12:45:47.000Z
---

> Index of every page on this site: [llms.txt](https://skilld.dev/llms.txt)

> **Skill from skilld.dev.** Follow the user's requested action.
> If the user asked to fork this Skill, follow the fork workflow below. Do not execute the Skill instructions while copying it.
> Otherwise, follow the instructions below for this session. You do not need to install anything.
>
> If the user asked to install this Skill, run `npx skilld install microsoft/scope/resilience`. Install writes the Skill files into the project, so every session loads them.
>
> ## Fork workflow
>
> A fork creates an editable local Skill with its original author and licence. The request authorizes copying and local installation.
> 1. Check `./skills/resilience`, the project lockfile, and selected Agent targets together. If the local directory or installed Skill exists, stop. Never overwrite an existing directory or Agent target.
> 2. Read [source metadata](https://skilld.dev/api/v1/skills/microsoft/scope/resilience) once. Use sourceUrl, sourceCommit, skillPath, sourceGone, and license. If the source is gone or its path is missing, stop. If license is null, read licence files at the source commit.
> 3. Fetch only the source commit into a temporary Git repository. Do not clone full history. Derive repository_url from sourceUrl, including repository renames. If sourceCommit is absent, resolve the sourceUrl ref once. Set source_commit to that actual commit. Run these commands in one shell call:
>
> ```sh
> git init --quiet "$temporary_dir"
> git -C "$temporary_dir" fetch --quiet --depth=1 "$repository_url" "$source_commit"
> git -C "$temporary_dir" checkout --quiet --detach FETCH_HEAD
> ```
>
> Read applicable licence declarations and notices at that commit. If copying is not permitted, report the restriction and stop.
> 4. Inspect source entries together, then copy the directory containing skillPath into `./skills/resilience`. Use the user's path if selected. Keep the original SKILL.md, relative links, scripts, binary assets, and executable modes. Exclude .git metadata. Reject symlinks and paths outside the Skill directory. After checking entries, use cp -a where available. A regular source directory needs no custom copy script. Do not save this page wrapper as SKILL.md.
> Preserve author credit, notices, and applicable licence files from repository or parent directories. Add PROVENANCE.md with the Skill page, source URL, actual commit, original path, and licence. Retain any existing PROVENANCE.md and record new provenance separately. Batch source inspection, copying, and provenance work where practical.
> 5. In the project root, run `skilld install ./skills/resilience --mode copy --plain`. If skilld is unavailable, use `npx skilld install ./skills/resilience --mode copy --plain`. This known command needs no help lookup. Install does not support --json. Use detected Agent targets, or add --agent for the targets the user selected. Install the local path, never the upstream selector. If installation fails, preserve the local copy and report the exact failure.
> 6. Confirm the local lockfile source and installed Agent copies once. Report the local path, actual commit, and Agent targets. After edits, reinstall the same local path. Upstream updates must not replace it. Do not publish or push unless the user asks.

# Resilience Patterns

When writing or reviewing code that makes outbound calls (REST APIs, database operations, queue interactions), always consider transient failure handling.

## Core Principle

**Every outbound HTTP call should have a retry strategy unless there's a specific reason not to.** Network blips, temporary 503s, and connection resets are normal in distributed systems. Code without retry is fragile by default.

## Available Tools

This repo provides retry utilities in `packages/shared/src/utils/retry.ts`:

### `@Retry` decorator (preferred for methods)

```typescript
import { Retry } from "shared";

class MyWorker {
  @Retry({ maxRetries: 3, baseDelayMs: 1000, isRetryable: () => true })
  async callExternalService(): Promise<void> {
    const res = await fetch("http://api/endpoint", { method: "POST" });
    if (!res.ok) throw new Error(`HTTP ${res.status}`);
  }
}
```

### `withRetry` function (for inline/conditional logic)

```typescript
import { withRetry } from "shared";

const result = await withRetry(
  () => fetch("http://api/data").then(r => r.json()),
  { maxRetries: 3, baseDelayMs: 500, isRetryable: (err) => isTransient(err) }
);
```

## Decision Checklist

When you encounter an outbound call, ask:

1. **Is this call idempotent?** If yes (GET, PUT, DELETE with idempotency key) -> retry aggressively.
2. **Is this call non-idempotent?** If yes (POST creating resources) -> retry cautiously, only on network errors (never got a response), not on 5xx (server may have processed it).
3. **Is this best-effort?** (e.g., triggering reports, sending notifications) -> retry with try/catch at call site so failures don't propagate.
4. **Is this critical path?** (e.g., saving user data) -> retry and propagate failure if exhausted.

## What to Retry

| Condition | Retry? | Reason |
|-----------|--------|--------|
| Network error (ECONNREFUSED, ETIMEDOUT) | Yes | Transient connectivity issue |
| HTTP 429 (Too Many Requests) | Yes | Rate limiting, respect Retry-After header |
| HTTP 500, 502, 503, 504 | Yes | Server-side transient failures |
| HTTP 400, 401, 403, 404, 409 | No | Client errors indicate a permanent problem |
| DNS resolution failure | Yes (limited) | May be temporary DNS propagation |

## Configuration Defaults

For service-to-service calls within the cluster:
- `maxRetries: 3`
- `baseDelayMs: 1000` (1 second)
- `maxDelayMs: 5000` (5 seconds)

For external API calls:
- `maxRetries: 5`
- `baseDelayMs: 500`
- `maxDelayMs: 30000` (respect rate limits)

## Anti-patterns

- **No retry at all** on service-to-service calls -- fragile
- **Retry on 4xx** -- wastes resources, will never succeed
- **Retry without backoff** -- can overwhelm a recovering service
- **Retry without a cap** -- can hang indefinitely
- **Retry non-idempotent POST without checking** -- can create duplicates
- **Stacking retry layers** -- never combine an HTTP client's built-in retry with `withRetry`/`@Retry` around the same call, or attempts multiply (`maxRetries × client.limit`). See below.

## Single retry layer

Retry belongs to **exactly one** layer per call. The `ky` HTTP clients used by the
CLI ([apps/cli/src/utils/api-client.ts](../../../apps/cli/src/utils/api-client.ts)) and Portal
([apps/portal/src/lib/api-client.ts](../../../apps/portal/src/lib/api-client.ts)) set `retry: 0`
precisely so they never stack with the shared cockatiel retry.

If a call needs transient-failure retries, pick **one** layer:

- **Enable the HTTP client's own `retry`** (with its own backoff), OR
- **Wrap the call in `withRetry` / `@Retry`** from `packages/shared/src/utils/retry.ts`.

Never both — combining them compounds attempts to `maxRetries × client.limit`.

## Reference

Full API documentation: [docs/architecture/retry.md](../../docs/architecture/retry.md)
