Exort Embedded Workspace

The README describes renderer access to native capabilities through a preload IPC boundary owned by the Electron main process, and maps workspace files, Arduino CLI, serial access, and local state. This supports some least-privilege and data-flow transparency, but the supplied material does not detail authorization boundaries or per-capability permissions, so neither earns full marks. The recommended workflow has users select a board and port and review edits in the editor, but does not say whether changes, compilation, or uploads require per-action confirmation; user confirmation earns 1. Test source covers managed-runtime integrity checks, path-traversal rejection, migration rollback, and error handling. Credential encryption, retention, and cleanup are not adequately described, and controls or recovery guidance for external effects—especially compilation and upload—are limited. Updater tests show restoration of runtime data and version plus cleanup of a failed candidate, supporting a 3 for rollback. The README, package metadata, and license identify the project and its license, but the author email is empty and no maintainer response or release responsibility is given, so attribution and maintenance responsibility are only partly clear.

The README's product description, feature list, and architecture diagram are broadly consistent, and package scripts expose build and typecheck commands; static material does not establish runtime consistency, so self-consistency earns 2. Dependency versions, Node/npm requirements, and Arduino CLI and OpenCode runtime requirements are described, and the release workflow shows install and packaging steps. The app still relies on external tools, model providers, and compatible board environments, with no availability guarantees, so dependency availability earns 2. Test source covers offline errors, version mismatches, and migration failures, while the updater distinguishes quiet background errors from actionable manual errors. This is source evidence, not an executed result, and there is no comprehensive user-facing error guide, so failure messages earn 2.

The README names learners, prototyping, and debugging as use cases and lists Arduino, ESP32, RP2040, STM32, and other board families, earning 3 for audience and scenarios. It says support depends on board cores, toolchains, USB drivers, and upload protocols and describes the desktop IPC boundary, but does not comprehensively enumerate Agent operations and limits, so capability boundaries earn 2. Trigger behavior is summarized as asking the Agent and following a recommended workflow; the material does not explain when tools run automatically or which actions require user initiation, so trigger precision earns 1. Desktop and source-run requirements, a Linux GPU startup issue, and board dependencies are documented, but platform-specific adaptation details are incomplete, so environment fit earns 2.

The README is organized around features, architecture, requirements, installation, workflow, and supported boards. It is clear but lacks a deeper configuration reference, so information architecture earns 2. Download and source-install steps, common npm commands, and some environment requirements are present, but full troubleshooting and configuration details are missing, so install notes earn 2. Repository and workspace names, version metadata, and commands appear consistent in the supplied material, but evidence is limited, so naming stability earns 2. The recommended workflow and command examples are concrete; no FAQ is provided, so examples and FAQ earn 2. Board support is qualified by local cores, toolchains, drivers, and protocols, and a Linux GPU issue is described; other known limitations are not comprehensively listed, so this earns 2. The README, package.json, and LICENSE all specify AGPL-3.0-only, and the license file is explicit, earning 3. package.json gives version 0.4.0 and the release workflow checks tags against the desktop package version and generates release notes, but no hand-maintained changelog appears in the supplied material, so versioning and changelog earn 2. The package author has a name but no email, and the material does not identify a maintenance team, support commitment, or accountable update path, so maintenance responsibility earns 1.

The product combines code editing, compile/upload, serial monitoring, and plotting, and the README gives a workflow for using them in embedded iteration. It does not provide interface or result examples that establish output quality, so output usability earns 2. Integrating an OpenCode coding agent with board management, compilation/upload, and live serial observation offers clear added value for embedded development, earning 3. Free models, bring-your-own providers, and local workspace history reduce some barriers, but users still need runtime, toolchain, and hardware setup; no cost or performance data is supplied, so cost-benefit earns 2.

The README makes locatable claims about features, requirements, and architecture, while test source gives concrete examples of runtime update, validation, and rollback logic. The supplied material does not map every product claim to implementation or verification records, so claim traceability earns 2. The README, package metadata, release workflow, and test source corroborate aspects of runtime management and release process. Some documented features lack corresponding source evidence here, and workflow configuration is not proof of an executed run, so cross-source corroboration earns 2. The README separates requirements, support conditions, and recommended steps, and test source exposes test scenarios; however, not all product claims are explicitly labeled by evidence status, and static material cannot establish that tests ran, so fact-inference separation earns 2.