.SYSTEMX 1.8.0-alpha.1 — Agent X and Agent Z

This alpha adds two standard roles alongside Agent 0: Agent X records human, bot, and automation events and due items; Agent Z uses a fixed, versioned set of 100 questions across 10 categories to produce evidence-backed scorecards and comparable deltas.

Agent X does not execute schedules. Agent Z does not call a model, run checks, approve deployment, or accept tasks. Agent 0/user retains coordination and acceptance. Repeated event keys, review inputs, and completed scorecards reuse existing records so unchanged work does not become another processing loop.

• Explicit roles-init preview and roles-init --apply activation for the selected root or .SYSTEMXP child; project policy, records, and memory are preserved.
• Versioned 10-by-10 review policy; pass/partial/fail/unknown/justified N/A answers, category scores, evidence coverage, immutable-by-command reports, and policy-aware comparisons.
• Workspace-target, path/link, JSON, manifest, configuration-error, interrupted-setup, record-size, and library integration fixes found during the runtime and lifecycle review.
• Updated public README and manual, including the complete rubric, commands, prompt examples, lifecycle guidance, and review/readiness limits.

Compatibility: New installations stay pinned with manual updates. Existing installations need an explicit update and role activation. Upgrade the external library to receive manager fixes; additive updates preserve existing bootstrap files. After activating X/Z IDs, downgrading to tools that predate these roles requires a reviewed record migration or a pre-activation backup.

Verification: 143 tests discovered in the local macOS run (138 passed, five filesystem-dependent skips), validated 126-file distribution, isolated wheel installation, preserved 1.7-to-1.8 upgrade through the old launcher, root/child isolation, replay reuse, and full uninstall/restore with X/Z records. The attached verification report records exact-commit CI results and artifact hashes.

Use at your own risk. This is alpha software and may change daily. Template checks and scorecards do not certify a downstream application's production readiness. The implementation is a local, cooperating-writer record system, not a distributed scheduler or event bus.

Agent X manual · Agent Z manual · 100 questions · Code review and readiness