Linux kernel: explicit AI guardrails and clear governance, but monolithic architecture and absent verification infrastructure make autonomous agent contribution impractical at scale.

The Linux kernel provides a comprehensive README with explicit role-based guidance including a dedicated section for AI Coding Assistants, plus architecture and build documentation, but lacks a machine-readable agent manifest (AGENTS.md/CLAUDE.md) with consolidated build/test/run commands.

Build process is documented but entirely manual, unversioned, and requires substantial human judgment to succeed reproducibly.

Linux kernel is a massive, ancient monolith with no conventional layout abstraction, no typed function signatures enforced by language, and no file-size constraints—agents lose context and edit wrong subsystems routinely.

Verification infrastructure is nearly absent; no CI, no runnable test suites, no type checking, no automated feedback loop — an agent cannot close its own loop.

Repository explicitly documents AI agent responsibilities (DCO certification, attribution, licensing), has a hardened .gitignore with no committed secrets, and threat model documentation; however, lacks a SECURITY.md contact path, CODEOWNERS boundaries, or .env.example patterns.

A 67,676-file monolithic kernel without build-output or node-modules segregation, linguist markers, or a distillable agent entry point; an LLM cannot efficiently navigate this without semantic indexing the entire codebase.

Exemplary first impression: the README immediately states purpose, scope, maturity, and audience with curated pathways for nine distinct user roles, backed by extensive internal documentation and a stable build system.

Explicit prerequisites and copy-pasteable build commands exist in detailed docs, but no lockfiles, version pins, or env templates; configuration is interactive and platform-dependent.

Excellent role-based navigation and deep documentation anchor navigation, but directory structure remains opaque and module boundaries are implicit rather than explicit.

Extensive process documentation exists but no dedicated CONTRIBUTING file, no issue/PR templates, no beginner-labeled issues, and contribution submission is entirely email-based mailing-list workflow — workable for insiders, rough for GitHub newcomers.

The Linux kernel is the gold standard of maintained open-source projects: continuous commits (as recently as 2026-06-21), subsystem maintainers with explicit SLAs across the entire codebase, rigorous documented contribution processes, active mailing-list-driven development, and thousands of collaborative contributors with clear governance.

Legally and socially safe: GPL-2.0 license is explicit and machine-readable, Code of Conduct is documented and enforced, and help channels (mailing lists, IRC, Bugzilla, MAINTAINERS) are clearly stated—only contribution templates are absent.