Documentation

Overview

AI / AGENT START HERE

Read AI Start Here and AI Provenance & Doc Sync before acting on the roadmap or creating commits.

If Architect says “read the GitHub”, “read the roadmap”, “get caught up”, or equivalent, run the canonical READ_ONLY_BOOTSTRAP first. Reconcile roadmap intent against RAG/history, local working trees, Git history, GitHub/submodules and relevant live evidence. Work only on the remaining delta; do not duplicate equivalent functionality.

Current documentation state: SYNCED as of 2026-08-07 (work ID XCSV-AI-001). The first SOVRAN-1 reconciliation is complete. The AI contract, provenance policy, tooling decision and GUARD development programme were absorbed into the desktop roadmap as Phase 15 by append, with no existing section rewritten; this wiki keeps its deliberate abstraction of the desktop’s measurements, RPT filenames and refuted hypotheses. That abstraction is the intended layering, not drift.

A dedicated Arma 3 Exile server on Tanoa, and every piece of software that keeps it alive.

This wiki is the operator/development map. The site is the public-facing view; the hub repo is the repository map.

Start here

page when you need it
AI Start Here mandatory AI navigation, reconciliation, delta-first workflow and GitHub work-tracking contract
AI Provenance & Doc Sync AI commit attribution, issue work receipts, and safe desktop/GitHub documentation reconciliation
AI / Development Tooling selected/rejected workflow tools, GitHub-native tracking decision, optional MCP and recommended CI/security helpers
XCSV GUARD Development Plan full current design programme: Gauntlet, restart safety, replay, backplane, UX, evidence, deployment, staging, player systems
Roadmap current order, completed work and planning priorities
Architecture understanding what talks to what
Runbook the server is broken and you need the operating procedure
Repositories where a given file lives and why
XM8 Apps player/admin-facing app work
Memory project memory / retrieval workflow
Lessons failures, refuted hypotheses and rules bought with real incidents

Workflow/tooling decision

XCSV uses GitHub Issues + GitHub Projects as the primary execution tracker so tasks, commits, PRs and sub-issues remain attached to the source instead of being copied into a second board. The official GitHub MCP Server is an optional enhancement after the baseline git/gh workflow is verified. Linear remains the only serious external future candidate, and only if GitHub Projects proves to have a measurable limitation.

See AI / Development Tooling for the full decision record.

Current GUARD programme

GUARD is being evolved around these principles:

Core invariants:

GUARD should never need to remember that the server is healthy. It should be able to prove the server is healthy again every time it starts.

A production failure should surprise XCSV once; afterward it becomes evidence, a fixture or an executable check.

Player-development direction

The estate is already large. Prefer coherence and progression over blind addon count: Territory Manager, contracts/jobs, bounties, meaningful faction standing, Chronicle/community systems, choreographed world events and selective objective systems. Keep heavyweight additional AI frameworks parked until measured headroom supports them.

Non-negotiables