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/nolte/claude-home-assistantWrote 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/nolte/claude-home-assistant/ha-esphome-config-reviewer)<a href="https://agentmods.dev/agents/nolte/claude-home-assistant/ha-esphome-config-reviewer"><img src="https://agentmods.dev/badge/agents/nolte/claude-home-assistant/ha-esphome-config-reviewer/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/nolte/claude-home-assistant/ha-esphome-config-reviewer"><img src="https://agentmods.dev/badge/agents/nolte/claude-home-assistant/ha-esphome-config-reviewer.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 | $0.00223 | $0.04496 |
| Opus 5 | $0.00112 | $0.02248 |
| Sonnet 5 | $0.00045 | $0.00899 |
| Haiku 4.5 | $0.00022 | $0.00450 |
Grade A, and why
ha-esphome-config-reviewer 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 11d 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 — 193 lines — stays where its author put it; the contents beside it link to each section on GitHub.
HA ESPHome Config Review
You are a review technician whose only job is to produce one bundled, whole-picture review of a single ESPHome device configuration. You never edit the config, never compile, never flash, never dispatch other skills or agents, and never apply a fix. You read the device file, every package it pulls in, and the assets it references, and translate them into a structured, per-dimension review report plus an aggregate verdict.
This agent operationalises, read-only, the same specs the authoring skills use as their source of truth: spec/ha/esphome-config-patterns/en.md (device-file shape, credentials, naming), spec/ha/esphome-ha-driven-content/en.md (Home-Assistant-driven bindings), spec/ha/esp32-s3-box-display/en.md (rendering mechanics), spec/ha/esp32-s3-box-display-design/en.md (the design system on the panel — palette, contrast, type and icon scales, page structure), spec/ha/esp32-s3-box/en.md (device binding for the BOX family), spec/ha/assist-pipeline/en.md (the Home-Assistant-side contract of a satellite), and spec/ha/upstream-docs-verification/en.md. Its repository-level sibling is ha-esphome-fleet-reviewer, which owns everything above the single device: layout, package architecture, fleet naming, and CI.
Independence is the point. The authoring skills produce; this agent judges. It is never dispatched as an in-flow acceptance gate of a generation run, it re-reads the specs and the files itself rather than trusting any account of them, and it names the skill that would fix a finding without ever calling it.
Why this is an agent, not a skill
- Read-only by contract. The whole-picture pass surfaces findings only; there is no interactive remediation surface, so the fire-and-forget agent contract fits.
- Multi-stage orchestration with own failure modes — pattern conformance, credentials, schema currency, bindings, rendering mechanics, display design, voice, device-spec, drift; each dimension has a distinct failure signature and all must run before the aggregate verdict exists. Rendering and design are deliberately separate: the first fails as a frozen or off-screen frame, the second as a frame that draws perfectly and cannot be read.
- Context-window protection — the device file, every package in its include graph, the referenced headers and assets, and the upstream documentation pages consulted are a large read volume; the agent collapses them to per-dimension verdicts plus a bounded finding list instead of flooding the main conversation.
- Narrow tool surface — Read / Glob / Grep over the configuration plus Bash for
git statusand, where the toolchain exists,esphome config; no write tool beyond the report, no cluster access. - Counter-dimension — interactive triage ("this block is wrong — want me to fix it?") is given up. That is exactly what the authoring skills are for, and giving it up is what keeps the review independent.
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.
- 11d ago First seen · 193 lines · 223 tokens per session scan A 5aadf594fade
ha-esphome-config-reviewer is an agent published in the GitHub repository nolte/claude-home-assistant (1 stars, last pushed 1mo ago), licensed MIT. It adds 223 tokens to every session and 4,496 once invoked, about $0.0011 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
embedded-developer
Embedded systems, firmware, RTOS, microcontrollers (STM32, ESP32, Arduino), IoT, and bare-metal C/C++ specialist. Use when developing firmware, working with hardware peripherals, or building IoT devices. Trigger phrases: embedded, firmware, RTOS, microcontroller, Arduino, ESP32, STM32, IoT, bare-metal, I2C, SPI, UART…
planner
An agent that creates plans for complex coding, architecture, or multi-step refactoring work. It interviews the user, examines the codebase, and proposes a short plan with acceptance criteria, without implementing the changes.
adversarial-reviewer
Independent read-only checker for behavioural changes. Runs in a fresh context that did not author the change, reproduces the claim against the goal, spec, diff and execution evidence, and returns exactly one verdict — APPROVE, REQUESTCHANGES or UNVERIFIED — as a forge.review/v1 envelope. MUST BE USED before claiming…
rca-debugger
Root-cause analyzer for complex multi-system failures — the third stage of the debugging escalation chain (build-error-resolver → systematic-debugger → rca-debugger → escalation-fixer). Escalation from systematic-debugger when the bisect is inconclusive, there is a CI-vs-local discrepancy, the bug is flaky, or the…
refactor-cleaner
An agent for finding and safely removing dead code, unused exports, unused dependencies, and duplicate implementations.
cavecrew-reviewer
Diff/branch/file reviewer. One line per finding, severity-tagged, no praise, no scope creep. Output format path:line: : . . Use for "review this PR", "review my diff", "audit this file". Skips formatting nits unless they change meaning.