Add observe-only monitor milestone for issue #4

[arena-ai-coding-agent]

Scope

Adds a bounded observe-only monitor milestone and safe development entrypoints. The collector uses bounded proc/sys reads, monotonic host sampling, explicit invalid/stale observations, stable device identity selection, single-instance ownership, and fixture-tested per-session PID/start-time, RSS, and transcript-motion primitives. A new foreground JSONL CLI exposes the host observer only; it does not auto-start, enumerate sessions, or enforce policy.

The follow-up adds:

• scripts/run.sh / scripts/dev.sh entrypoints, just monitor, and Debug/ReleaseSafe CLI build/help/version coverage.
• Real just check, just lint, just test, and just quality recipes.
• Fail-closed unsupported root build/install/start/E2E/benchmark paths; just clean preserves tracked build/ sources.
• Claude/Gemini/Copilot instructions pointing to shared conventions; Claude session startup is read-only and installs no packages.
• A corrected aspect scan preserving the owner-only SPDX warning.

No hooks, signals, process writes, LMDB integration, or memory balloon were added. Existing licensing declarations and issue #2 are unchanged.

Verification

• 29 core Zig test blocks pass in Debug and ReleaseSafe; the monitor CLI compiles and its help/version paths are exercised in both modes.
• just check, just lint, and just quality pass locally. Quality reports the existing missing SPDX header in src/interface/ffi/src/main.zig as an owner-only warning; it was not altered.
• just doctor, just --list, shell syntax checks, zig fmt --check, and ShellCheck on changed shell scripts pass.
• just build, just install, and just e2e fail explicitly rather than claim success; just clean preserves build/.
• A live paired run on this Linux sandbox returned invalid:missing for the initial baseline and a valid idle_tell observation on the next sample. Lock exclusion and rejection of symlink/permissive lock files and a symlink runtime directory were exercised. This is one host, not topology acceptance.

Not a closure claim

Issue #4 is not ready to close. Paired accuracy across supported hosts/storage topologies, managed session discovery/startup identity capture, versioned schema/retention/IPC, actual OFF propagation through an integration, resource-pressure/failure acceptance, and host/VM durability evidence remain open. No balloon or global hook was installed. scripts/run.sh is a developer dispatcher, not a conformant desktop/service launcher: there is no application runtime, so service start and integration modes fail. Full launcher-standard conformance is deferred until a real application exists. Issue #2 remains owner-only/open.

Closes no issue; references #4 for incremental review.

[coderabbitai]

Important

Review skipped

Bot user detected.

To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Advanced

Run ID: 58d75547-9083-4310-b1c8-7236ca65b98e

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}