# Reliability Patterns

- kind: Skill
- sku: sku_reliability_patterns
- version: 1.0.0
- price: $29 USD
- availability: in_stock
- runtime: OpenClaw, Claude Code, Codex, Grok
- last_tested: 2026-08-12
- page: https://kitsforagents.com/kits/skills/reliability-patterns/
- json: https://kitsforagents.com/kits/skills/reliability-patterns.json

## Job

Keep a small agent job from failing quietly: fixed output shape, isolated fetches, missed-run backup, logged failures, labeled keys, output limits, independent health check.

## Who it is for

You already have a briefing or cron running and are tired of noticing the break a day late.

## Who should skip it

You do not have an agent installed yet. Start with SOUL.md if the rules are still fuzzy.

## What you get

Installable skill folder with the seven reliability patterns Aaron uses on a Raspberry Pi OpenClaw setup: fixed output templates, per-fetch error wrappers, a backup heartbeat, labeled .env keys, hard output limits, logged failures, and an independent health-check job. Includes eval fixtures for a missed cron and a single dead API.

## Inputs

- existing_agent_dir (path, required): Path to the current skills or config folder
- delivery_surface (enum): telegram, local, or none

## Outputs

- skill_folder (skill_md): Drop-in reliability skill with health-check script
- checklist (markdown): Seven patterns marked present or missing

## Eval

86/100 · 6/7 · 2026-08-12 · Raspberry Pi 5 OpenClaw + Telegram

Morning briefing still delivered when a news API returned 503. Missed 6:40 cron recovered by the backup check. Consecutive-failure alerting was not fully wired.

## Success criteria

- Health check exists as its own scheduled job, not inside the briefing skill
- A single failed fetch marks that section unavailable and still completes
- Output stays under the advertised word cap on three consecutive fixture runs

## License and refund

Personal-use, one seat, no redistribution. 14-day refund if the advertised eval fails on the advertised fixture, or the checksum mismatches.

## Checkout

Human: POST https://kitsforagents.com/api/v1/checkout with `{ "sku": "sku_reliability_patterns" }`
Agent: POST https://kitsforagents.com/api/acp/checkout_sessions with ACP 2026-04-17 checkout session body.

```yaml
schema: "makelky.agent-packet/2026-08"
sku: "sku_reliability_patterns"
name: "Reliability Patterns"
kind: "skills"
version: "1.0.0"
url: "https://kitsforagents.com/kits/skills/reliability-patterns/"
md_url: "https://kitsforagents.com/kits/skills/reliability-patterns.md"
json_url: "https://kitsforagents.com/kits/skills/reliability-patterns.json"
job: >
    Keep a small agent job from failing quietly: fixed output shape, isolated fetches, missed-run backup, logged failures, labeled keys, output limits, independent health check.
summary: >
    The seven patterns from a Raspberry Pi that actually broke. A skill folder plus a health-check so a missed 6:40 cron does not cost the whole day.
inputs:
  - name: "existing_agent_dir"
    type: "path"
    required: true
    description: "Path to the current skills or config folder"
  - name: "delivery_surface"
    type: "enum"
    required: false
    description: "telegram, local, or none"
outputs:
  - name: "skill_folder"
    type: "skill_md"
    description: "Drop-in reliability skill with health-check script"
  - name: "checklist"
    type: "markdown"
    description: "Seven patterns marked present or missing"
runtime:
  primary: "openclaw"
  also:
    - "claude-code"
    - "codex"
    - "grok"
permissions: "filesystem write-skills-dir; network limited for health checks"
secrets_required: []
eval:
  score: 86
  casesPassed: 6
  casesTotal: 7
  dateTested: "2026-08-12"
  environment: "Raspberry Pi 5 OpenClaw + Telegram"
  summary: >
      Morning briefing still delivered when a news API returned 503. Missed 6:40 cron recovered by the backup check. Consecutive-failure alerting was not fully wired.
price:
  amount: 29
  amount_minor: 2900
  currency: "USD"
  display: "$29"
  is_digital: true
  availability: "in_stock"
license:
  id: "personal-use"
  url: "https://kitsforagents.com/legal/license/"
  redistribution: "forbidden"
  seats: 1
  commercial: false
install:
  after_purchase: true
  preview_url: "https://kitsforagents.com/install/reliability-patterns/"
success_criteria:
  - "Health check exists as its own scheduled job, not inside the briefing skill"
  - "A single failed fetch marks that section unavailable and still completes"
  - "Output stays under the advertised word cap on three consecutive fixture runs"
refund:
  window_days: 14
  policy_url: "https://kitsforagents.com/refunds/"
  agent_endpoint: "https://kitsforagents.com/api/v1/refunds"
  trigger: "success_criteria_unmet_or_install_checksum_mismatch"
budget:
  class: "mid"
risk:
  level: "low"
author:
  name: "Aaron Makelky"
  url: "https://aaronmakelky.com/about"
  tested_on_own_work: true
  brand_deals: false
checkout:
  human: "https://kitsforagents.com/api/v1/checkout"
  acp: "https://kitsforagents.com/api/acp/checkout_sessions"
last_tested: "2026-08-12"
is_eligible_checkout: true
```