raydthanh/aihostcheck

AIHostCheck solves a practical problem for AI-assisted development: agents often suggest commands without knowing the user’s operating system, installed tools, hardware, or shell. The project provides that context through a small, dependency-free Go CLI with human-readable and versioned JSON output. Its evidence model is especially thoughtful. Capabilities can report unknown, unsupported, error, or permission_denied instead of turning every failed check into a false absence claim. The implementation supports Windows, macOS, and Linux through native collectors and keeps command execution behind one controlled boundary with fixed arguments, timeouts, output limits, and home-directory redaction. Privacy and security receive real attention through offline operation, excluded identifiers, private vulnerability reporting, JSON schemas, compatibility rules, and clear warnings about sharing reports. CI builds and tests on all three operating systems, while the release workflow creates six platform archives, verifies installation on native runners, publishes checksums, and adds provenance attestations. The Apache 2.0 license, contribution guide, changelog, architecture notes, and release documentation make the repository unusually complete for an early project.
The main limitation is the gap between automated platform testing and real-device validation. Most tests confirm report structure, basic probe behavior, and schema rules. They do not fully exercise the many operating-system commands, parsing paths, permission failures, unusual installations, or hardware combinations that determine whether the results are accurate. The repository openly acknowledges this and currently includes only limited physical-host evidence, including a Windows AMD64 GPU timeout verification. Expanding the validation matrix across Windows, macOS, Linux, ARM64, multiple GPUs, containers, and restricted environments should be the highest priority. Installation also remains less friendly than it could be because users must manually download archives, verify checksums, and handle unsigned binaries. The documented decision to delay package managers and OS signing is reasonable, but those additions would improve adoption when maintenance capacity allows. Overall, AIHostCheck is a focused, carefully documented tool with strong privacy boundaries and a credible engineering foundation. Broader real-world testing would turn its sound design into more dependable evidence for AI agents.