Skill Claude CodeCodex
Default workflow for coding tasks in this repo — scope with the user, build/test in a warm agent-N worktree, open a PR, iterate on review, then merge and reset on explicit approval. Use for any new implementation task, not only when the user names a slot or worktree explicitly.
Skill Claude CodeCodex
Hand a decision brief, design, or risky diff to a fresh second-opinion agent (codex CLI, an Opus/Fable subagent, or another session) for design feedback or a targeted adversarial pass, and route the results back through the fix-ladder triage. Use when the user asks for a codex/second-opinion/design review or a consult…
Skill Claude CodeCodex
Record footage of any game situation, with native gizmo diagnostics drawn over it, and hand the user a clip. Use when footage is the deliverable — showing a behavior, visually diagnosing a sim bug, demoing a feature, or recording RL episodes.
Skill Claude CodeCodex
Coordinate access to this repository's shared Unity editors. Use before running Unity tests, opening an interactive Unity editor, driving a live editor through the unity CLI, or diagnosing why another agent cannot use Unity.
Instructions file CodexOpenCode
AGENTS.md instructions for amindell11/astronomical-home, covering agents.md, fix ladder, scope conservation, comments and vocabulary.
Instructions file
Claude Code instructions for amindell11/astronomical-home: Claude Code entry shim — the canonical always-loaded core is AGENTS.md, imported below. Edit that file, not this one; the two entry points stay identical by construction.
Agent
This game is designed in an Obsidian vault outside the repo. Treat these two notes as primary context — for design/doc work, and for knowing what to build next.
Agent
Issues live in this repo's GitHub Issues; use the gh CLI. The repo is public — every issue body and comment is world-readable.
Agent
This repo is backed by a persistent, file-based agent memory — durable facts, decisions, and the why/how behind tracker issues. It lives outside the repo, in the primary session's memory directory.
Agent
Review law for scripts/. Read it before changing a script's outputs or calling one from another script. Provenance: doc/FeaturePlans/ScriptSeamHardening.md.
Agent
explicit outDir/-OutDir).
Agent
Repo-side companion to the machine-generated unity-cli skill (/.claude/skills/unity-cli/, rendered from the binary by unity skill refresh — never edit it; repo deltas live here). The editor-side surface is experimental (com.unity.pipeline) on a beta CLI: re-verify the bugs below after any version bump. Coordination…
Agent
Camera.main never run in Start/OnEnable/Initialize/update loops or any runtime path. Cache in Awake; Initialize(...) only assigns injected references (see AGENTS.md → dependency wiring).