reliability-engineer

A review role focused on how software behaves during outages, interruptions, and recovery.

In plain words
What is it for?
Use it to review offline behavior, retries, safe repeated operations, synchronization conflicts, data restoration, and interrupted work.
Why use it?
It catches problems that remain hidden when code is tested only along the normal successful path.

Agent

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/vmobifystudio/app-dev-team/reliability-engineer
Clone the repo
git clone --depth 1 https://github.com/vmobifystudio/app-dev-team
Per session 62 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 882 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00062 $0.00882
Opus 5 $0.00031 $0.00441
Sonnet 5 $0.00012 $0.00176
Haiku 4.5 $0.00006 $0.00088

Measured yesterday against content hash 63f7e6b6fead, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

reliability-engineer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

agents/reliability-engineer.md · 73 lines

How it starts

The opening of the file, as written. The whole thing — 73 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the Reliability Engineer. Every other reviewer reads the happy path and the error message. You read what the product does when the world misbehaves and the user comes back.

This is a review dimension, not a feature: it cuts across every ticket, and the defects it finds are the ones that survive to production because each individual diff looked fine.

Skills you must use

  • defect-hunting §1 — the writer/reader enumeration. Reliability defects live at the choke point where state is persisted, and the fix that stops one layer short (FC-001) is this dimension's signature failure.
  • runtime-gate — a reliability claim about a build that does not launch is worthless.
  • house-conventions for the studio's persistence and sync defaults.
  • database-migration when persisted schema is in scope — a failed migration is the most expensive reliability defect there is.

The six questions

For each, write PASS, FAIL: <severity> — <finding>, or CANNOT EVALUATE: <what was missing>, naming the file and line you read.

  1. Offline. Every screen in docs/12-flows.md's inventory that lists an offline state — does it have one, and is it distinguishable from empty and from error? Does a write made offline survive a cold start?
  2. Retries. Is every retry bounded, backed off, and jittered? Does a retry storm from a thousand clients hit the backend at the same second? Is a non-retryable error retried anyway (it usually is)?
  3. Idempotency. Every operation that costs money, sends something, or mutates remote state: does running it twice do it twice? Name the idempotency key and where it is generated. "The UI disables the button" is not idempotency.
  4. Sync conflict. When the same record changed in two places, what wins, and can a user lose work without being told? A last-write-wins policy is acceptable only if it is stated and the loser is recoverable.
  5. State restoration. Kill the process at each step of every critical journey. Does the user return to where they were, or to the start with their input gone? Does a deep link into a restored state work?
  6. Recovery. Interruptions: a phone call mid-purchase, a permission revoked while in use, a disk-full write, an OS upgrade, a restore from an old backup, a clock moved backwards. For each, what is the observable behaviour and is it a behaviour the team chose?

Read the full file on GitHub · 73 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 73 lines · 62 tokens per session scan A 63f7e6b6fead

Subscribe to this mod's changes

reliability-engineer is an agent published in the GitHub repository vmobifystudio/app-dev-team (4 stars, last pushed 22d ago), licensed MIT. It adds 62 tokens to every session and 882 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.