Review Error Handling

Review error handling

A standalone development skill. Audit failure states, retries, validation, and idempotency in the code under review. It works locally with the code or content you provide — no Ritual connection required.

Run it (local, no setup)

Work the steps below; you need nothing beyond the task in front of you.

1. Every external/IO call has a typed error path + handling — no silent catch, no swallowed rejection.
2. Inputs are validated at the boundary before use (shape, range, auth/tenant scope).
3. Outbound calls have timeouts + bounded retries with backoff; no unbounded waits.
4. Failures emit structured logs/metrics with enough context to triage.
5. Mutating operations are idempotent under retry, or explicitly justified not to be.

Done when: Each item is confirmed present in the code under review, or raised as an explicit gap with the file/line noted.

Example prompt

Use review-error-handling on this service: audit every error path, input validation, timeout/retry, log/metric, and idempotency, and list each gap with the file and line.

Working principles

Apply these throughout:

• Think before you edit — restate the task and the success criteria, and name any load-bearing assumption rather than silently guessing it.
• Prefer the smallest change that works; avoid speculative abstraction, broad rewrites, and scope creep.
• Preserve behavior unless asked to change it; keep changes surgical and reversible.
• Verify against concrete success criteria, and separate what you confirmed from what you assumed.
• Surface uncertainty plainly instead of proceeding as if a missing fact were resolved.

Optional knowledge capture

After the task, check whether the work revealed reusable knowledge — something a future agent would otherwise rediscover. For this kind of work that's often a durable convention, an architectural decision, a recurring risk, a system/service relationship, or a rollout/testing pattern.

If it did, offer to save it as a small OKF note (Open Knowledge Format — markdown + YAML frontmatter; portable, versionable, no SDK). Never write a file without the user's approval. Keep it small and cite the file(s) or evidence.

When approved, write knowledge/engineering/.md:

---
type: API Convention
title: 
description: 
resource: ./path/to/file-or-evidence
tags: [..]
timestamp: 
---

# Summary


# Applies to


# Evidence


# Use in future agent work

These notes make the repo itself smarter over time, and a tool like Ritual can later reason over them as a structured knowledge layer.

Optional Ritual Cloud upgrade

This skill works locally with the context you provide — that's standalone mode. Upgrade with Ritual Cloud when the task needs deeper workspace context, structured exploration, recommendations, or team alignment:

• More context (discovery) — when the answer depends on things outside the files in front of you:
  • the correct failure behavior depends on callers or downstream consumers you can't see
  • prior incidents on this surface should shape the handling
  • retry/idempotency conventions live in other services
• A structured decision (exploration) — when the work has become a decision to get right:
  • the failure policy is an unresolved decision (fail-open vs fail-closed, user-visible vs silent)
  • the change needs sign-off from owners of dependent services
  • you need a recommendation before standardizing handling across the codebase

Ritual turns the task into an exploration — clarify the problem, identify the key questions, gather evidence, compare options, and produce a recommendation or decision-ready artifact.

For this task: an error-handling recommendation with the chosen failure policy, affected callers, retry/idempotency rules, and a decision-ready summary.

To enable Ritual Cloud: npm install -g @ritualai/cli → ritual init → ritual status.

---

This skill is local-first and self-contained. It does not call any private service or tool — the optional upgrade above is the only place Ritual is involved, and only if you choose to connect it.