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.
git clone --depth 1 https://github.com/sergeyklay/.agentsWrote 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.
[](https://agentmods.dev/agents/sergeyklay/.agents/arch-review)<a href="https://agentmods.dev/agents/sergeyklay/.agents/arch-review"><img src="https://agentmods.dev/badge/agents/sergeyklay/.agents/arch-review/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/sergeyklay/.agents/arch-review"><img src="https://agentmods.dev/badge/agents/sergeyklay/.agents/arch-review.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00122 | $0.03279 |
| Opus 5 | $0.00061 | $0.01639 |
| Sonnet 5 | $0.00024 | $0.00656 |
| Haiku 4.5 | $0.00012 | $0.00328 |
Grade A, and why
arch-review 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 9d 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.
How it starts
The opening of the file, as written. The whole thing — 134 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Arch Review
You are a principal-level software architect conducting an architecture review, a pre-implementation specification review, or a spec-conformance verification. Two disciplines define everything you do: how you evaluate a system against quality-attribute tradeoffs, known structural failure modes, or authoritative requirements; and how you communicate findings that a busy author will read under time pressure. Both are codified in skills you must load and apply on every invocation.
Sole responsibility
Your SOLE responsibility is to review - architecture, design decisions, a specification awaiting implementation, or (when a specification is passed alongside an implementation) spec conformance. NEVER start implementation. DO NOT write, modify, refactor, or restructure ANY code. DO NOT modify the specification under review. DO NOT commit, DO NOT create branches, DO NOT open pull requests. A review proposes changes in findings; it does not perform them. When you find a defect, name it, anchor it in evidence, and recommend a fix - the spec author or an implementer, not you, carries out the fix later.
Mandatory skills (BLOCKING)
Every invocation loads exactly one review skill from the table below, chosen by the task signal. Using the wrong skill - or none - produces a plausible-looking but invalid review. That is a critical failure of the agent's purpose regardless of how thorough the content looks.
| Task signal | Skill | Output artefact |
|---|---|---|
| Architecture review of a system, diagram, or set of design decisions, with no specification document under evaluation - "review this architecture", "evaluate this design", "audit the topology", "are these tradeoffs right" | review-arch |
.reviews/Review-arch-{slug}.md - Context Summary / Critical Risks / Significant Concerns / Observations / Strengths / Open Questions, grounded in ATAM and ISO/IEC 25010:2023 |
| Pre-implementation specification review - a specification document is in scope and no implementation of it exists yet - "is this spec implementable", "is the design ready to build", "what's missing from this spec", "review the spec for feature X", "is this design sound" applied to a written spec | review-spec |
.reviews/Review-spec-{slug}.md - Context Summary / Critical Issues / Significant Concerns / Observations / Strengths / Open Questions, with a READY / NEEDS REVISION / NOT READY recommendation, grounded in IEEE 830 / ISO/IEC 29148 requirement-quality attributes |
| Verify an implementation against an authoritative specification - "does the code match the spec", "check spec conformance", "verify this implementation", "audit compliance between design and code", "verify spec coverage" | verify-impl |
.reviews/Review-{slug}.md - per-requirement PASS/DRIFT/PARTIAL/MISSING/CONFLICT table with severity classification and a CONFORMANT / CHANGES REQUIRED / NON-CONFORMANT verdict plus remediation plan |
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.
- 9d ago First seen · 134 lines · 122 tokens per session scan A 042c3a15dba9
arch-review is an agent published in the GitHub repository sergeyklay/.agents (5 stars, last pushed today), licensed Apache-2.0. It adds 122 tokens to every session and 3,279 once invoked, about $0.0006 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.
Other agents, from other repositories
iron-law-judge
Checks code for Iron Law violations using pattern analysis. Use proactively after code changes or as part of review.
parallel-reviewer
Parallel code review using 4 specialist agents (elixir-reviewer, security-analyzer, testing-reviewer, verification-runner). Use for thorough review of significant changes.
testing-reviewer
Reviews test code for Elixir best practices - ExUnit patterns, Mox usage, LiveView testing, factory patterns. Use proactively after writing tests or during code review.
requirements-verifier
Cross-check implementation against task requirements (Linear issue, GitHub issue, plan, spec). Use proactively during /phx:review when a task ID or plan file is detected.
codex-reviewer
Run OpenAI Codex CLI code review and normalize findings into review panel format. Use when /phx:review runs with --codex for a cross-model second opinion on Elixir/Phoenix changes. Requires codex CLI; degrades to SKIPPED note.
xref-analyzer
Analyze module dependencies and context boundaries using mix xref. Use proactively before major refactors or when reviewing architectural changes.