Pi/OMP/OpenAI/Codex/Antigravity entry. Canonical context: .agents/instructions.md.
Read it first. It points to always-on rules + on-demand refs. If the active agent does not natively expand standalone
@path references, it must use its read tool to load every referenced file and apply this rule recursively.
Commands in .agents/commands/:
add-pip-dep.md- add pip depbump-exts-changelog.md- bump ext versions/changelogscommit.md- commit with repo stylecreate-branch.md- create feature branchcreate-extension.md- scaffold extensioncrash-debug.md- gather/classify RTX Remix Toolkit crash evidencedebug-extension-load.md- debug load/test discoverykit-test.md- run/debug Kit testsprepare-mr.md- prep MRremix-review.md- scripted RTX Remix code reviewremix-review-follow-up.md- verify and resolve reviewed threadsremix-review-post-threads.md- post reviewed findings as inline threadsremove-extension.md- remove extension + refs
Skills in .agents/skills/:
add-pip-dep, agent-config, bump-exts-changelog, commit, completion-gates, create-branch,
create-extension, crash-debug, debug-extension-load, documentation, extension-docs, kit-test,
memory-promotion, prepare-mr, remix-review, remix-review-follow-up, remix-review-post-threads,
remove-extension, rtx-remix-modding.
Auto-use completion-gates before done. Auto-use memory-promotion when durable project knowledge appears.
Internal checkout? If .agents/commands/internal/README.md exists, read it for internal commands.
Subagents:
docs->.agents/subagents/docs.mdunit-test-writer->.agents/subagents/unit-tests.mde2e-test-writer->.agents/subagents/e2e-tests.mdusd-expert->.agents/subagents/usd-expert.mdui-expert->.agents/subagents/ui-expert.mdreviewer->.agents/subagents/review.md(scripted immutable packets only; whole reviews useremix-review)