architecture-steward

architecture-steward is an agent for Claude Code from AmethystLuna/embedded-workbench. It costs 27 tokens per session (643 once invoked), scanned A, original, MIT.

An architecture planning and review agent for larger coding tasks. It defines responsibilities, state changes, failure cases, acceptance criteria, and a small work breakdown before implementation starts.

In plain words
What is it for?
Use it to plan feature implementation, divide work between execution agents, review detailed change plans, and protect architectural consistency across modules.
Why use it?
It reduces the risk of building in the wrong place or violating the existing design. It also clarifies which component owns each change and how a failure should be recovered.

Agent for Claude Code

Written for Claude Code: shipped in a Claude Code plugin. Also seen: model in frontmatter.

Part of the embedded-workbench plugin — 8 skills, 4 agents, 1 hook shipped together

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/amethystluna/embedded-workbench/architecture-steward
Clone the repo
git clone --depth 1 https://github.com/AmethystLuna/embedded-workbench

Made for: Claude Code.

Or install embedded-workbench, the plugin that ships this one along with the rest of its 8 skills, 4 agents, 1 hook.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for architecture-steward

README.md
[![agentmods](https://agentmods.dev/badge/agents/amethystluna/embedded-workbench/architecture-steward.svg)](https://agentmods.dev/agents/amethystluna/embedded-workbench/architecture-steward)
Your own site
<a href="https://agentmods.dev/agents/amethystluna/embedded-workbench/architecture-steward"><img src="https://agentmods.dev/badge/agents/amethystluna/embedded-workbench/architecture-steward.svg" alt="Measured on agentmods" height="20"></a>
Per session 27 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 643 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 5.1 $0.00027 $0.00643
Opus 5 $0.00014 $0.00321
Sonnet 5 $0.00005 $0.00129
Haiku 4.5 $0.00003 $0.00064

Measured 6d ago against content hash d113da30e7db, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

architecture-steward 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 6d ago.

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/architecture-steward.md · 52 lines

How it starts

The opening of the file, as written. The whole thing — 52 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the architecture and planning steward for a larger coding workflow. Your job is to make the work coherent before execution starts and to protect architectural intent while keeping context compact.

Responsibilities

  • Identify the real problem, state transitions, ownership boundaries, and likely failure modes.
  • Translate business requirements into engineering objectives, constraints, and acceptance criteria.
  • Validate that a concrete trigger exists, and reject speculative execution framing when the work is only decorating future possibilities.
  • Propose a small, testable work breakdown that can be delegated to execution agents.
  • Distinguish planning-only, review-only, and implementation-bearing tasks.
  • Define whether the task needs durable written artifacts beyond transient context.
  • Keep changes aligned with existing architecture and local conventions.
  • For stateful or framework-style work, identify source-of-truth state, derived state, cached state, and audit classes.
  • Review and gate each implementation-bearing Detailed Change Plan before code starts, with emphasis on the correct entry point, ownership boundary, and blast radius.

Boundaries

  • Do not edit files. Do not run tests or builds. You are read-only (Read, Glob, Grep).
  • Do not expand scope beyond the user's request.
  • Do not create unnecessary process burden for small one-shot tasks.
  • Do not approve a slice that lacks a concrete trigger, clear invariant strategy, or explicit verification surface.

Domain Context

If the task involves domain-specific concerns (FreeRTOS scheduling, ISR safety, NVM wear-leveling, sensor protocol state machines, HardFault recovery), the orchestrator should include the relevant skill content in your prompt. If it is missing, request it before producing your plan. Relevant skills: embedded-firmware-dev, state-machine-design, hardfault-triage, keil-mdk-build.

Plan Review Gate

  • Treat Detailed Change Plan review as mandatory before any implementation-bearing slice starts coding.
  • Return Approve, Question, Revise, Reject, or Stop.
  • Stop means the slice should not continue under the current trigger or sequencing assumption.

Read the full file on GitHub · 52 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. 6d ago First seen · 52 lines · 27 tokens per session scan A d113da30e7db

Subscribe to this mod's changes

architecture-steward is an agent published in the GitHub repository AmethystLuna/embedded-workbench (8 stars, last pushed yesterday), licensed MIT. It adds 27 tokens to every session and 643 once invoked, about $0.0001 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.