docs: README hook, discoverability, Pages site, game-behaviour reference

- README: lead with what it is + the AI-agent-collaboration research
  framing + current status in the first paragraph; rename H1 to
  "GoldenEye 007 PC Port"; add a live CI badge.
- New "How this differs from the other GoldenEye PC projects" section:
  this is a source port of the N64 original; the Recomp projects
  binary-translate the Xbox 360 XBLA remaster (different game, no shared
  code).
- CITATION.cff for GitHub's "Cite this repository" widget / Zenodo.
- GitHub Pages site off /docs (_config.yml + index.md landing page).
- agentic-development.md: searchable title + abstract.
- .github/social-preview.{png,html,md}: repo social-preview card styled
  after the GE mission-briefing screen, kept editable (CSS source + a
  regen/how-to doc beside the PNG).
- docs/dev/game-behavior-reference.md: distilled the raw GameFAQs
  walkthrough dump down to the behaviour-defining parts (combat/AI model,
  difficulty scaling, per-level objective sets, timers, weapon data,
  original-game quirks). Credits: Brett "Nemesis" Franklin, marshmallow.

Claude-Session: https://claude.ai/code/session_01GULf6knbsqfsnW39tMr9WT
This commit is contained in:
James Dansereau committed 2026-09-02 00:27:05 -04:00
1 parent 0c2673f636
commit 4e2b721ac8
9 files changed
+707 -17

No files matched your search

+8 -6
View File
@@ -1,10 +1,12 @@
# Agentic development: a two-agent experiment
# Agentic development: two AI coding agents porting GoldenEye 007
*A GoldenEye 007 N64→desktop port, taken from "builds a ROM" to "playable front
end" in about two weeks — driven mostly by two coding agents, a local
open-weight model and a hosted frontier model, handing work back and forth
through shared written notes under one person's direction. This is the goal,
the method, the timeline, and an honest read on what did and didn't work.*
*A case study in AI-agent collaboration on a large, low-level codebase. A
GoldenEye 007 Nintendo 64 → desktop port, taken from "builds a ROM" to
"playable front end" in about two weeks — driven mostly by two coding agents (a
local open-weight model, Qwen 3, on a single RTX 5090, and a hosted frontier
model, Claude / Claude Code), handing work back and forth through shared
written notes under one person's part-time direction. This is the goal, the
method, the timeline, and an honest read on what did and didn't work.*
## Contents