linxisa/linx-isa/linx-omx
Claude Code and agent skill by linxisa/linx-isa
- community
Checking that the SKILL.md file is still in the repository.
Use when operating inside /Users/zhoubot/linx-isa with OMX and you need LinxISA-specific task routing, gate triage, leaf-before-repin discipline, strict closure, dual-lane convergence, or evidence-backed bring-up workflows.
Install
npx vskill@latest install linxisa/linx-isa/linx-omxSKILL.md
Linx OMX
Use this skill when the task is about using OMX well inside the LinxISA superproject.
This skill does not replace the domain skills such as linx-superproject, linx-compiler, linx-qemu, linx-linux, linx-core, linx-pycircuit, or linx-lib. It routes OMX behavior so those domain skills are used in the right order and with the right proof requirements.
Core Rules
- Resolve ISA questions from
isa/v0.58/linxisa-v0.58.jsonand the v0.58
generated/reference surfaces. Never route normative work to historical profiles or archived narrative.
- Route Tile-engine questions through the exact
VEC,SFU,TLSU, and
CUBE taxonomy. VEC is element-wise only and SFU owns complex operations. TEPL is the unchanged Mode/Function encoding carrier for VEC/SFU, not an engine; aliases do not allocate new opcode space.
- Treat the superproject as the coordination plane, not the place to hide leaf
fixes.
- Start from the smallest failing contract, gate, or test.
- Trust
docs/bringup/gates/latest.jsonover markdown summaries. - Run topology and static policy checks before trusting runtime closure.
- Fix leaf modules first; repin second; run strict closure third.
- Keep evidence path clean: command, lane, run-id, SHA manifest, pass/fail, and
artifact path.
- Do not run
avs/qemusuites in parallel against the same output directory.
Default Startup
From /Users/zhoubot/linx-isa:
Before architecture work, verify the current contract directly:
Use --madmax only when you intentionally want the less constrained path.
OMX Routing
Use $deep-interview when:
- the failing gate owner is unclear,
- module boundaries are unclear,
- lane policy is unclear,
- stop conditions are unclear.
Prompt shape:
Use $ralplan when:
- scope is known but gate order is still fuzzy,
- you need to lock repin policy,
- you need to define exact closeout proof,
- multiple plausible owners exist and the integration boundary must be fixed.
Prompt shape:
Execute directly when:
- one owner and one module are already obvious,
- the task is small and reversible,
- tmux/worktree durability is not needed.
Use native subagents when:
- you have several small read-only questions,
- the next step benefits from parallel lookup,
- the work does not need durable tmux or worktree state.
Use omx team when:
- the task is long-running and cross-owner,
- tmux/worktree durability matters,
- you need manifest-aligned lanes such as
arch,llvm,qemu,linux,
libc, linxcore, pycircuit, or integration.
Do not use omx team for trivial fanout.
Use $ralph or omx ralph when:
- the plan is already approved,
- one owner should keep pushing until verification is complete,
- the task benefits from a persistent completion loop.
Mandatory Checks
Topology preflight
Static policy guard
PR-tier closure
Dual-lane runtime convergence
Domain Handoff
Route the work into the owning domain skill after OMX routing is settled:
linx-superprojectfor cross-repo closure, topology, repin, and evidence
packaging.
linx-compilerfor compile AVS and LLVM closure.linx-qemufor runtime AVS, strict system, opcode sync, and emulator issues.linx-linuxfor Linux boot and initramfs closure.linx-corefor RTL, block structure, testbench, and trace-adjacent RTL work.linx-pycircuitfor model-diff and interface-contract work.linx-libfor glibc/musl closure.linx-isafor architecture contracts and docs gates.
Evidence Contract
For every serious run, leave behind:
- exact command,
- lane,
- run-id,
- SHA manifest,
- pass/fail result,
- artifact or log path under
docs/bringup/gates/.
If you cannot point to the evidence path, the run is not complete.
Anti-Patterns
Do not:
- trust
GATE_STATUS.mdoverlatest.json, - repin before the owning leaf proof is green,
- debug dual-lane parity before leaf correctness,
- route work to forbidden legacy paths,
- run shared-output runtime suites in parallel,
- use
omx teamwhere native subagents or direct execution are enough.
References
Read these as needed:
/Users/zhoubot/linx-isa/docs/project/omx-linxisa-playbook.md/Users/zhoubot/linx-isa/docs/project/omx-linxisa-prompt-templates.md/Users/zhoubot/linx-isa/docs/project/superproject-bringup-methodology.md/Users/zhoubot/linx-isa/docs/project/new-agent-sop.md/Users/zhoubot/linx-isa/docs/project/maintainer-repin-checklist.md/Users/zhoubot/linx-isa/docs/bringup/agent_runs/manifest.yaml
Security
- Security scanPassed, 100 of 100Pattern checks on the skill's files.
- Model reviewNot runA model review is scheduled for the next rescan.
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/linx-omx/SKILL.md
Popularity
- Repository stars
- 82
- Repository forks
- 16
- 7-day trend score
- 22
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/linxisa/linx-isa/linx-omx)