tetsuo 7etsuo

Ashpit character and graphics design document

Status: production outline for the first full art pass
Source of truth checked: current Ashpit gameplay code on 2026-08-27
Scope: characters, animation, world art, items, effects, interface art, and the runtime data needed to display them
Out of scope: sound, music, voice, and new gameplay rules

What this document covers

Ashpit currently renders the arena, players, enemies, shots, pickups, hazards, and effects with colored shapes. The gameplay code already defines the classes, enemy roles, bosses, weapons, upgrades, relics, contracts, objectives, and run options. This document turns those mechanics into an art production list.

AgenC memory system — implementation handoff

Status: implementation-grade plan; no production implementation has been performed by this document.

Research and repository audit date: 2026-07-22 (America/Edmonton).

Primary decision: build AgenC's memory governance and execution-state control plane natively in TypeScript. Treat third-party memory projects as optional, replaceable adapters or sources of tested ideas—not as security or correctness authorities.

How to use this handoff

/goal — Ship the best possible AgenC marketplace

LOCAL OPERATING LEDGER — NEVER COMMIT, PUSH, PR, OR MERGE THIS FILE. This lowercase TODO.md is deliberately separate from the tracked historical TODO.MD. It is excluded through .git/info/exclude. Keep it local and self-update it throughout the goal.

Goal

Build, verify, release, and operate AgenC as exceptionally clean, secure,

REFACTOR v5 — AgenC Modular Runtime Architecture

OS-inspired layered architecture. Thin kernel, typed contracts, everything is a plugin.

v5 — Ground-up redesign based on deep research into OS kernels (Linux LKM, seL4, QNX), plugin systems (VS Code, webpack Tapable, Grafana, Home Assistant), agent frameworks (LangGraph, OpenAI Agents SDK, Anthropic Agent SDK, AgentForge, Auton), and full coupling analysis of the actual codebase.

1. What's Wrong with v4

CHAT

Recommendation

The best research-backed fit for AgenC is:

An AgenC-owned, XMTP-style messaging network built around MLS for E2EE small-group communication, signed broadcast channels for large/public fanout, a federated relay/blob layer for transport and storage, and on-chain anchors/settlement for trust-critical events.

In practice, that means:

	#!/usr/bin/env bash

	set -Eeuo pipefail

	updates_api="https://api2.cursor.sh/updates"
	downloads_origin="https://downloads.cursor.com"
	profile_source="/usr/share/apparmor/extra-profiles/bwrap-userns-restrict"
	profile_target="/etc/apparmor.d/bwrap-userns-restrict"

	fail() {

	You are setting up my automated schedule system in Grok Bot. Two Grok Bot
	teammates: the Scheduler plans my days into Google Calendar, and SARGE
	checks the evidence and keeps me on plan. You will drive the Grok Bot
	desktop app over the Chrome DevTools Protocol, mine 90 days of my local
	activity, create both bots, and fill my calendar. Work through the nine
	steps in order. Stop and ask me whenever a step needs a choice, an
	approval, or a sign-in. Hard rules for the whole run: never send email,
	invites, or messages to anyone outside this system (talking to me and to
	the two bots inside Grok Bot is fine); only create, move, or delete
	calendar events whose titles start with [TB]; never type passwords or

	main-BKrkOjLl.js:12 noa-engine v0.33.0
	thinInstanceMesh-EsmKqdEl.js:4 BJS - [06:30:54]: Babylon.js v6.49.0 - WebGL2 - Parallel shader compilation
	main-BKrkOjLl.js:2 MaxListenersExceededWarning: Possible EventEmitter memory leak detected. 11 beforeRender listeners added. Use emitter.setMaxListeners() to increase limit
	at c (main-BKrkOjLl.js2⃣4469)
	at Wf.r.addListener (main-BKrkOjLl.js:2⃣752)
	at new jy (main-BKrkOjLl.js:402:7732)
	at main-BKrkOjLl.js:483:16913
	i @ main-BKrkOjLl.js:2
	main-BKrkOjLl.js:2 MaxListenersExceededWarning: Possible EventEmitter memory leak detected. 11 tick listeners added. Use emitter.setMaxListeners() to increase limit
	at c (main-BKrkOjLl.js:22⃣69)

	┌─────────────────────────┐ ┌──────────────────────────────────┐
	│ agenc-backend │ mints │ LiteLLM gateway box (separate) │
	│ (CONTROL PLANE) │ ──key──▶ │ (DATA PLANE) │
	│ │ │ │
	│ • login (wallet) │ │ • holds the REAL Anthropic / │
	│ • billing (Stripe/Sol) │ │ OpenAI / xAI / Google keys │
	│ • mints per-user keys │ │ • routes + enforces budget │
	│ • picks model │ │ • tracks spend (its own DB) │
	│ │ │ • egress locked to provider hosts │
	│ NEVER sees provider │ └──────────────▲────────────────────┘

	ANGLES: 4 SOURCE CHECKS: 13
	======================================================================
	SUMMARY: De-slopping LLM text is mostly a structural and thinking problem, not a punctuation problem. Naive negative prompts ("don't use em dashes," "sound less like AI," blanket word blacklists) fail for two reasons: the model has no replacement target so it substitutes another tell, and the deeper issue is generic, thinking-free scaffolding that survives any word swap. What works is a staged process: fix structure first (kill default intros/recaps, symmetric subheads, list-dumps, signposting), then apply positive language rules that say what to write instead, then a separate judge-model pass that only diagnoses, then a human pass for argument and voice. The most reliable single heuristic is the "one-line test": compress each paragraph to one sentence and read the result as an outline; if it reads like a generic template, no word-level edit will save it. Editors uniformly recommend editing in passes (one tell