Borrowing it
Nothing to install: this file belongs to maoxx241/vllm-ascend-workspace. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/maoxx241/vllm-ascend-workspace/main/.agents/skills/ascend-triton-kernel-optimization/SKILL.mdgit clone --depth 1 https://github.com/maoxx241/vllm-ascend-workspaceWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/maoxx241/vllm-ascend-workspace/ascend-triton-kernel-optimization)<a href="https://agentmods.dev/skills/maoxx241/vllm-ascend-workspace/ascend-triton-kernel-optimization"><img src="https://agentmods.dev/badge/skills/maoxx241/vllm-ascend-workspace/ascend-triton-kernel-optimization/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/maoxx241/vllm-ascend-workspace/ascend-triton-kernel-optimization"><img src="https://agentmods.dev/badge/skills/maoxx241/vllm-ascend-workspace/ascend-triton-kernel-optimization.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00125 | $0.00680 |
| Opus 5 | $0.00063 | $0.00340 |
| Sonnet 5 | $0.00025 | $0.00136 |
| Haiku 4.5 | $0.00013 | $0.00068 |
Grade A, and why
ascend-triton-kernel-optimization scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 45 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Ascend Triton Kernel Optimization
Optimize one correct kernel against a comparable NPU baseline. Preserve correctness and evidence integrity at every round.
Workflow
- Require a passed
ascend-triton-kernel-validationmanifest for the exact candidate hash and case set. - Record target SoC/software versions, NPU baseline, warmup/repeat policy, per-case medians, noise threshold, case-regression limit, and performance objective.
- Query
.agents/knowledge/with current profiler signals and failure signatures before selecting an optimization. - Run
scripts/triton_optimization.py planto lock the baseline, objective, cases, and starting kernel. - Use
msprof opor the target version's supported operator profiler to compare device time, block count, MTE2/MTE3, Vector, Scalar/FLOWCTRL, and pipeline gaps against theoretical floors. - Form one falsifiable hypothesis, change one primary mechanism, and rerun the full correctness gate on the candidate.
- Normalize per-case measurements and run
record. The controller choosesKEEP,DISCARD,NOISE, orFAIL; never overwrite the best kernel. - After repeated failures, enter diagnosis and change the bottleneck model rather than retrying the same parameter.
- Run
analyzeonly when the objective is met, round budget is exhausted, or the run intentionally stops.
Entry point
scripts/triton_optimization.py provides:
plan: validate correctness evidence and lock baseline/configuration;record: enforce sequential rounds, parent hash, full-case validation, measurement coverage, and noise-aware decisions;analyze: report best kernel, cumulative improvement, discarded hypotheses, and terminal objective status.
Read:
- Behavior contract for config, round, decision, and status semantics.
- Profiling decision tree before choosing a hypothesis.
- Ascend optimization techniques only for the diagnosed bottleneck.
- Command recipes for the lifecycle.
- Acceptance before claiming an optimization.
What ships with it
8 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- agents/openai.yaml 258 B
- references/acceptance.md 1.1 KB
- references/ascend-techniques.md 2.8 KB
- references/behavior.md 2.9 KB
- references/command-recipes.md 973 B
- references/profiling-decision-tree.md 2.3 KB
- scripts/triton_optimization.py 20 KB runs code
- tests/test_triton_optimization.py 6.6 KB runs code
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago First seen · 45 lines · 125 tokens per session scan A 4cabf6f4b4bd
ascend-triton-kernel-optimization is a skill published in the GitHub repository maoxx241/vllm-ascend-workspace (36 stars, last pushed 6d ago), licensed MIT. It adds 125 tokens to every session and 680 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
gke-compute-classes
Configures, optimizes, and troubleshoots GKE ComputeClasses. Use when configuring Spot VMs with on-demand fallback, targeting specific accelerators (GPUs/TPUs) or machine families, restricting ComputeClass access, or debugging pending pods related to node pool auto-creation. Do not use for cluster-level Node Auto…
jetson-diagnostic
Read-only Jetson health snapshot for identity, memory, GPU, thermal, power, storage, services, and top processes.
doca-socket-relay
Use this skill when the operator is driving the DOCA Socket Relay to bridge a socket-oriented host application onto a BlueField DPU peer without rewriting it — picking the deployment shape (in-process, sidecar, or BlueField service container), configuring the host-side socket and the DPU-side forwarding endpoint…
offensive-z-wave
Z-Wave attack methodology — sniffing with Z-Force / EZ-Wave / RTL-SDR + ZniffMobile, S0 (legacy) network-key derivation flaw and key reuse, S2 (modern) ECDH commissioning analysis, replay/injection on unauthenticated nodes, default-key brute-force on test deployments, and home-automation hub pivots. Use when targeting…
hsb-flash
Flash the FPGA on an HSB board connected to an NVIDIA devkit. Supports HSB Lattice boards (FPGA versions 2407, 2412, 2507, 2510) and Leopard Imaging VB1940 "all-in-one" cameras (FPGA versions 2507, 2510). Uses release-specific YAML manifests and board-type-specific program commands. Lattice and VB1940 commands must…
jetson-validate-image
Use after jetson-flash-image to run static BSP checks, on-target smoke/regression tests on a flashed DUT, or both. Not for build or flash steps. Triggers: validate bsp, on-target validation.