Add sourceosctl reasoning validation and inspection commands

[mdheller]

Purpose

Superconscious now emits deterministic governed reasoning artifacts, and SourceOS-Linux/sourceos-spec has canonical reasoning contracts for ReasoningRun, ReasoningEvent, ReasoningReceipt, ReasoningReplayPlan, and ReasoningBenchmark.

sourceosctl should expose operator-facing commands to validate, inspect, and replay-plan these artifacts.

Scope

Add guarded read-only commands:

sourceosctl reasoning validate <run-dir>
sourceosctl reasoning inspect <run-dir>
sourceosctl reasoning replay-plan <run-dir>
sourceosctl reasoning events <run-dir>

Initial implementation can wrap or port the Superconscious M1 validator and inspector semantics.

Acceptance criteria

• Commands are read-only.
• Commands require no network, model provider, credentials, browser automation, or host mutation.
• validate fails closed when required artifacts are missing, benchmark fails, replay plan is absent, or safe trace posture is invalid.
• inspect renders task, status, safe trace posture, coordination records, replay class, benchmark result, and event timeline.
• Documentation makes clear that sourceosctl inspects artifacts and does not become the cognition loop.

Related repos

• SocioProphet/superconscious
• SourceOS-Linux/sourceos-spec
• SocioProphet/agentplane
• SocioProphet/sociosphere