Kernel Verification
ZJLi2013/awesome-kernel-skills
5-stage kernel correctness verification protocol for Triton and CUDA kernels.
A skill your agent uses when you need a NEW or optimized MPK CUDA kernel (a per-task .cuh under include/mirage/persistentkernel/tasks/) that must PROVABLY beat a target WITHOUT the kernel-optimizer…
Install Ferret Kernel System
$ npx skills add mirage-project/mirage --skill ferret-kernel-system -a claude-codeProject install by default; add -g for ~/.claude/skills/.
$ gh skill install mirage-project/mirage ferret-kernel-system --agent claude-codeProject scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).
$ git clone --depth 1 https://github.com/mirage-project/mirage.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.claude/skills/ferret-kernel-system .claude/skills/ferret-kernel-system && rm -rf skills-srcUse ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.
Claude Code skills documentation · loads skills from .claude/skills/
Install the "ferret-kernel-system" agent skill from https://github.com/mirage-project/mirage/tree/mpk/.claude/skills/ferret-kernel-system into .claude/skills/ferret-kernel-system/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "ferret-kernel-system", then confirm the skill loads.Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$skill-installer install https://github.com/mirage-project/mirage/tree/mpk/.claude/skills/ferret-kernel-systemType this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.
$ npx skills add mirage-project/mirage --skill ferret-kernel-system -a codexProject install goes to .agents/skills/; add -g for ~/.codex/skills/.
$ gh skill install mirage-project/mirage ferret-kernel-system --agent codexProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/mirage-project/mirage.git skills-src && mkdir -p .agents/skills && cp -r skills-src/.claude/skills/ferret-kernel-system .agents/skills/ferret-kernel-system && rm -rf skills-srcUse ~/.agents/skills/ instead of .agents/skills for a personal install.
Codex skills documentation · loads skills from .agents/skills/
Install the "ferret-kernel-system" agent skill from https://github.com/mirage-project/mirage/tree/mpk/.claude/skills/ferret-kernel-system into .agents/skills/ferret-kernel-system/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "ferret-kernel-system", then confirm the skill loads.Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add mirage-project/mirage --skill ferret-kernel-system -a cursorProject install goes to .agents/skills/; add -g for ~/.cursor/skills/.
$ gh skill install mirage-project/mirage ferret-kernel-system --agent cursorProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/mirage-project/mirage.git skills-src && mkdir -p .cursor/skills && cp -r skills-src/.claude/skills/ferret-kernel-system .cursor/skills/ferret-kernel-system && rm -rf skills-srcUse ~/.cursor/skills/ instead of .cursor/skills for a personal install.
Cursor skills documentation · loads skills from .cursor/skills/, .agents/skills/, .claude/skills/, .codex/skills/
Install the "ferret-kernel-system" agent skill from https://github.com/mirage-project/mirage/tree/mpk/.claude/skills/ferret-kernel-system into .cursor/skills/ferret-kernel-system/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "ferret-kernel-system", then confirm the skill loads.Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gemini skills install https://github.com/mirage-project/mirage.git --path .claude/skills/ferret-kernel-system--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.
$ npx skills add mirage-project/mirage --skill ferret-kernel-system -a gemini-cliProject install goes to .agents/skills/; add -g for ~/.gemini/skills/.
$ gh skill install mirage-project/mirage ferret-kernel-system --agent gemini-cliProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/mirage-project/mirage.git skills-src && mkdir -p .gemini/skills && cp -r skills-src/.claude/skills/ferret-kernel-system .gemini/skills/ferret-kernel-system && rm -rf skills-srcUse ~/.gemini/skills/ instead of .gemini/skills for a personal install, then run /skills reload.
Gemini CLI skills documentation · loads skills from .gemini/skills/, .agents/skills/
Install the "ferret-kernel-system" agent skill from https://github.com/mirage-project/mirage/tree/mpk/.claude/skills/ferret-kernel-system into .gemini/skills/ferret-kernel-system/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "ferret-kernel-system", then confirm the skill loads.Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gh skill install mirage-project/mirage ferret-kernel-systemInstalls for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).
$ npx skills add mirage-project/mirage --skill ferret-kernel-system -a github-copilotProject install goes to .agents/skills/; add -g for ~/.copilot/skills/.
$ git clone --depth 1 https://github.com/mirage-project/mirage.git skills-src && mkdir -p .github/skills && cp -r skills-src/.claude/skills/ferret-kernel-system .github/skills/ferret-kernel-system && rm -rf skills-srcUse ~/.copilot/skills/ instead of .github/skills for a personal install. Commit .github/skills so cloud agent and code review can use it.
GitHub Copilot skills documentation · loads skills from .github/skills/, .claude/skills/, .agents/skills/
Install the "ferret-kernel-system" agent skill from https://github.com/mirage-project/mirage/tree/mpk/.claude/skills/ferret-kernel-system into .github/skills/ferret-kernel-system/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "ferret-kernel-system", then confirm the skill loads.GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add mirage-project/mirage --skill ferret-kernel-system -a opencodeOpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.
$ gh skill install mirage-project/mirage ferret-kernel-system --agent opencodeProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/mirage-project/mirage.git skills-src && mkdir -p .opencode/skills && cp -r skills-src/.claude/skills/ferret-kernel-system .opencode/skills/ferret-kernel-system && rm -rf skills-srcUse ~/.config/opencode/skills/ instead of .opencode/skills for a personal install.
OpenCode skills documentation · loads skills from .opencode/skills/, .claude/skills/, .agents/skills/
Install the "ferret-kernel-system" agent skill from https://github.com/mirage-project/mirage/tree/mpk/.claude/skills/ferret-kernel-system into .opencode/skills/ferret-kernel-system/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "ferret-kernel-system", then confirm the skill loads.OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
Facts
- Skill name
ferret-kernel-system- GitHub stars
- 2.5k
- Token cost
- ~1.3k tokens
- SKILL.md length
- 566 words
- Files
- 2
- Skills in repo
- 24
- Repo updated
- First seen
- Licence
- Apache-2.0
At a glance
A skill your agent uses when you need a NEW or optimized MPK CUDA kernel (a per-task .cuh under include/mirage/persistentkernel/tasks/) that must PROVABLY beat a target WITHOUT the kernel-optimizer…
- Works in 6 steps: TARGET — the exact MPK op + what it… → REAL-MATH CONTRACT — enumerate EVERY… → SHAPES — exact (M,K,N) at the real TP/EP… → …
- — phrasings like dispatch ferret to optimize a kernel
- SKILL.md covers The problem it solves (read…, The 3 agents (in…, How to invoke (the main-thread… and The invariants that make it…, plus 2 more sections
- Calls claude
What it does
Ferret Kernel System is an agent skill from mirage-project/mirage. Use when you need a NEW or optimized MPK CUDA kernel (a per-task .cuh under include/mirage/persistentkernel/tasks/) that must PROVABLY beat a target WITHOUT the kernel-optimizer cheating its own correctness test. This is the frozen-gate Ferret system — a 3-agent orchestration (dispatcher → independent test-writer → optimizer) that makes "the optimizer ships a simplified-math kernel and self-reports a passing test" structurally impossible. Triggers — phrasings like "dispatch ferret to optimize a kernel", "have an…
Its SKILL.md is about 1.3k tokens, which your agent loads only when the skill is triggered. The skill folder holds 1 other file (for example `DESIGN.md`).
It sits in Agent Workflows, covering GPU and accelerator computing and Multi-agent orchestration. It works with CUDA. The repository describes itself as: Mirage Persistent Kernel: Compiling LLMs into a MegaKernel. The licence is Apache-2.0.
When your agent uses it
- — phrasings like dispatch ferret to optimize a kernel
- Have an agent write a kernel but guarantee it doesnt simplify the math
- Generate a kernel the frozen-gate way: the user/main-thread wants a kernel optimized
- Generated by an agent
Example prompts
- “structurally impossible. Triggers — phrasings like”
- “have an agent write a kernel but guarantee it doesn”
- “generate a kernel the frozen-gate way”
- “/ferret-kernel-system”
Workflow steps
6 steps, taken from the first numbered list in SKILL.md.
- TARGET — the exact MPK op + what it replaces; the baseline = the kernel being
- REAL-MATH CONTRACT — enumerate EVERY step the kernel must compute, NO
- SHAPES — exact (M,K,N) at the real TP/EP regime (derive from the builder/weights).
- PRODUCTION COMPILE FLAGS — -rdc=true / MPK_FORCE_RDC_TRUE=1, arch sm_100a,
- ABI — the device task_impl signature + NS/NE.
- CANONICAL REFERENCE SOURCE — the already-trusted oracle to compare against
What it can do on your machine
Read from SKILL.md and the folder at commit f9eb70c. It shows what the files ask for, not the result of running them.
Tool permissions
Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.
From allowed-tools in the SKILL.md frontmatter.
Runs code
Shell commands in SKILL.md call:
claude
From the folder's file list and the shell code blocks in SKILL.md.
Network
No URLs in SKILL.md.
From URLs in SKILL.md, links to its own repository left out.
Credentials
Names no API keys, tokens, secrets or passwords.
From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.
Context cost
Ferret Kernel System loads about 1.3k tokens when it runs. Until then it costs about 202 tokens; SKILL.md has 566 words of instructions outside code blocks.
Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.
Safety
Auto-check passed
The automated check found no risky patterns in SKILL.md.
Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.
SKILL.md
The full file from mirage-project/mirage at commit f9eb70c, republished under its Apache-2.0 licence (© mirage-project). 566 words, ~1,266 tokens.
Download SKILL.mdSave it as
.claude/skills/ferret-kernel-system/SKILL.md (or your agent's skills folder). This skill also uses 1 other file; get the full folder from GitHub.- name
- ferret-kernel-system
- description
- Use when you need a NEW or optimized MPK CUDA kernel (a per-task `.cuh` under include/mirage/persistent_kernel/tasks/) that must PROVABLY beat a target WITHOUT the kernel-optimizer cheating its own correctness test. This is the frozen-gate Ferret system — a 3-agent orchestration (dispatcher → independent test-writer → optimizer) that makes "the optimizer ships a simplified-math kernel and self-reports a passing test" structurally impossible. Triggers — phrasings like "dispatch ferret to optimize a kernel", "have an agent write a kernel but guarantee it doesn't simplify the math", "generate a kernel the frozen-gate way": the user/main-thread wants a kernel optimized or generated by an agent, especially when correctness/no-simplification matters more than fire-and-forget speed.
Ferret frozen-gate kernel system — how to invoke + why
The problem it solves (read first)
The OLD ferret wrote its OWN correctness test. It once shipped a DeepSeek-V3 attention kernel with SIMPLIFIED math (theta-10000 rope not YaRN, 1/sqrt(576) not the YaRN mscale scale, head-sum o_proj skipping W_UV, no kv_a_layernorm) and self-reported cosine 1.0 — against its own simplified reference. Marking its own homework. The refactor takes judging + constraints OUT of the optimizer.
The 3 agents (in .claude/agents/)
| Agent | Role | You invoke? |
|---|---|---|
| ferret-kernel-agent (L1 dispatcher) | The ENTRY POINT. Pins ALL constraints, freezes the gate via the test-writer, runs the optimizer in-session, drives Codex review each round, does in-MPK faithful acceptance. | YES — this is the only one you invoke directly. |
| ferret-test-writer (L2a) | Writes the FROZEN, hash-locked gate vs a CANONICAL reference (never re-derived), checking INTERMEDIATE tensors. Spawned by L1, before the optimizer. | No (L1 spawns it) |
| ferret-optimizer (L2b) | In-session optimizer (replaces claude -p), judged ONLY by the frozen gate, can't simplify. Spawned by L1 each round. | No (L1 spawns it) |
How to invoke (the main-thread → subagent contract)
Invoke the ferret-kernel-agent subagent (via the Agent tool) with a COMPLETE constraint contract — this is where nothing gets missed and no simplification slips in. Give it, explicitly:
- TARGET — the exact MPK op + what it replaces; the baseline = the kernel being replaced, benched the way MPK calls it (not an external SOTA unless it's the consumer).
- REAL-MATH CONTRACT — enumerate EVERY step the kernel must compute, NO simplification (the test-writer turns this into intermediate checks).
- SHAPES — exact (M,K,N) at the real TP/EP regime (derive from the builder/weights).
- PRODUCTION COMPILE FLAGS —
-rdc=true/MPK_FORCE_RDC_TRUE=1, arch sm_100a, single-stream / no-CUDA-graph / no-cta_group::2. State that FINAL acceptance is the in-MPK faithful build; standalone-rdc=trueis diagnostic only. - ABI — the
__device__ task_implsignature + NS/NE. - CANONICAL REFERENCE SOURCE — the already-trusted oracle to compare against (the in-MPK task-chain output; the official HF model; an in-tree faithful test). NEVER "let the agent derive it." If the gate already exists (hash-locked in the workspace), tell L1 to reuse it (hash-verify) and go straight to the optimizer loop.
Show full SKILL.md (223 more words)Show less
The invariants that make it trustworthy (Codex-hardened — gate fidelity is load-bearing)
- The gate is built by the independent test-writer, not the optimizer.
- The reference is CANONICAL (validated against a trusted source), never re-derived.
- The gate checks INTERMEDIATES (golden vectors per stage), not just a final cosine
— a deep simplification (a dropped layernorm, a wrong rope base) is caught at the
first diverging stage (
first_failing_stage), not washed out. - Multiple metrics + edge cases (long-context, boundary positions).
- The gate is sha256 hash-locked; L1 re-verifies the hash before EVERY round (tamper = abort).
- The optimizer is judged ONLY by
gate/check.py. Codex reviews each round on two axes — Integrity (did it simplify to pass?) + Plan (is the lever sound?). - FINAL acceptance = in-MPK faithful build (
MPK_FORCE_RDC_TRUE=1, compiled into the real megakernel). A standalone number never ships. - Early stop = round incomplete, never success.
Proof it works
Re-run on the simplified attention: the optimizer (judged by the frozen gate) built a
CORRECT real-DSv3 fused attention in round 1 — GATE_RESULT {pass:true} on all 5
cases / all intermediates, gate untouched, dump provenance verified (W_UV BMM present,
real mscale²/sqrt(192) scale), all 4 simplifications fixed, Codex Integrity+Plan PASS.
The old simplified-kernel failure is now structurally impossible.
Reference
Full design + the /cd-mechanics + the hardening rationale: DESIGN.md (bundled
alongside this skill). The ferret runtime itself lives at
~/ferret/ (its CLAUDE.md is the optimizer's methodology). Pre-authored task specs:
~/ferret/tasks/.
© mirage-project, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file
Files
SKILL.md and 1 other file in .claude/skills/ferret-kernel-system of mirage-project/mirage.
- SKILL.md
- DESIGN.md
Open the folder on GitHubat commit f9eb70c
Compare with similar skills
Ferret Kernel System next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.
| Skill | Stars | Used in | Tokens | Auto-check | Licence | Repo updated |
|---|---|---|---|---|---|---|
| Ferret Kernel System this skillmirage-project/mirage | 2.5k | — | ~1.3k | Automated safety check: Pass | Apache-2.0 | |
| Kernel VerificationZJLi2013/awesome-kernel-skills | 102 | — | ~702 | Automated safety check: Pass | None | |
| Paddle Design CompilerPaddlePaddle/Paddle | 24k | — | ~3.6k | Automated safety check: Pass | Apache-2.0 | |
| Debug Distributed Hangsgl-project/sglang | 37k | 2 repos | ~2.4k | Automated safety check: Pass | Apache-2.0 | |
| CUTLASS FMHA Incremental Rebuildmicrosoft/onnxruntime | 22k | — | ~1.3k | Automated safety check: Pass | MIT | |
| Cuda Index Widthpytorch/pytorch | 104k | — | ~1.6k | Automated safety check: Pass | Custom licence |
Similar skills
Paddle Design Compiler
PaddlePaddle/Paddle
A skill your agent uses when working with Paddle 3.0 compiler full pipeline: SOT (Symbolic Opcode Translator) for bytecode-level dy2st graph capture, PIR (Paddle IR) for SSA-based intermediate…
Debug Distributed Hang
sgl-project/sglang
Debug hanging issues in SGLang distributed inference (TP/PP/DP/EP).
- Official
CUTLASS FMHA Incremental Rebuild
microsoft/onnxruntime
Explains why editing CUTLASS fused-MHA headers in ONNX Runtime can leave stale CUDA kernels after an incremental build, and how to force and verify a real rebuild.
Cuda Index Width
pytorch/pytorch
Choose 32-bit vs 64-bit index math in PyTorch CUDA kernels. An agent skill from pytorch/pytorch.
Graphsignal
graphsignal/graphsignal
Profile AI inference workloads (vLLM, SGLang, TensorRT-LLM, PyTorch, any GPU application) with the Graphsignal profiler and read the results from its local /signals JSON endpoint.
More from mirage-project/mirage
All 24 skills in this repoV2 Perf Iteration
mirage-project/mirage
Runtime-V2 performance-iteration workflow. An agent skill from mirage-project/mirage.
Add Mpk Task
mirage-project/mirage
Step-by-step guide for adding a new task implementation to Mirage Persistent Kernel (MPK).
B200 Flash Attention4 Planner
mirage-project/mirage
A skill your agent uses when the user wants to design or extend a FlashAttention-style forward kernel on B200/Blackwell, involving the two MMAs QKᵀ and PV, online softmax, S/P/O in TMEM, warp roles…
Mpk Faithful Gate
mirage-project/mirage
Build or run a FAITHFUL in-MPK per-task latency gate (slowCTA at the production grid + cos) for a DeepSeek-V3 MPK decode kernel or shape.
Mpk Lever Cleanup
mirage-project/mirage
A skill your agent uses when a batch of env-gated (ifdef MPKDSV3 / os.environ-controlled, default-OFF) MPK optimization levers needs to be consolidated into a single clean code path for a PR…
Test Mode
mirage-project/mirage
Guide for using MPK test mode to unit-test individual layers or multi-layer pipelines through the full compilation pipeline.
Related
Works with
Categories
Questions about Ferret Kernel System
What does Ferret Kernel System do?
A skill your agent uses when you need a NEW or optimized MPK CUDA kernel (a per-task .cuh under include/mirage/persistentkernel/tasks/) that must PROVABLY beat a target WITHOUT the kernel-optimizer…. Ferret Kernel System is an agent skill from mirage-project/mirage.cuh under include/mirage/persistentkernel/tasks/) that must PROVABLY beat a target WITHOUT the kernel-optimizer cheating its own correctness test.
When should I use Ferret Kernel System?
Ferret Kernel System fits situations like: — phrasings like dispatch ferret to optimize a kernel; have an agent write a kernel but guarantee it doesnt simplify the math; generate a kernel the frozen-gate way: the user/main-thread wants a kernel optimized; generated by an agent.
How do I install Ferret Kernel System in Claude Code?
Run `npx skills add mirage-project/mirage --skill ferret-kernel-system -a claude-code`. Or copy the skill folder (.claude/skills/ferret-kernel-system in mirage-project/mirage) into .claude/skills/ferret-kernel-system in your project. Claude Code loads it when a task matches its description.
How do I install Ferret Kernel System in Codex?
Run `npx skills add mirage-project/mirage --skill ferret-kernel-system -a codex`. Or copy the skill folder (.claude/skills/ferret-kernel-system in mirage-project/mirage) into .agents/skills/ferret-kernel-system in your project. Codex loads it when a task matches its description.
Can I use Ferret Kernel System in Cursor, Gemini CLI or GitHub Copilot?
Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add mirage-project/mirage --skill ferret-kernel-system -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/ferret-kernel-system, .gemini/skills/ferret-kernel-system, .github/skills/ferret-kernel-system and .opencode/skills/ferret-kernel-system in your project.
What does Ferret Kernel System need to run?
Going by SKILL.md and its folder, Ferret Kernel System needs the command-line tools its instructions call (claude).
Does Ferret Kernel System access the network?
SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.
Is Ferret Kernel System safe to install?
Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.
What licence does Ferret Kernel System use?
Ferret Kernel System is published under the Apache-2.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.
How many tokens does Ferret Kernel System use?
About 1.3k tokens (SKILL.md is roughly 5.1k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.
What are the alternatives to Ferret Kernel System?
Skills that share tags, products or a category with Ferret Kernel System: Kernel Verification (ZJLi2013/awesome-kernel-skills, 102 stars), Paddle Design Compiler (PaddlePaddle/Paddle, 24k stars), Debug Distributed Hang (sgl-project/sglang, 37k stars) and CUTLASS FMHA Incremental Rebuild (microsoft/onnxruntime, 22k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.
Who maintains Ferret Kernel System?
mirage-project (a GitHub organization) maintains it in mirage-project/mirage, which has 2,541 GitHub stars. The repository holds 24 skills in this directory. The repository was last updated on October 7, 2026.
Source: mirage-project/mirage on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.