Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add tensormux/kernel-skills --skill write-backend-agnostic-kernel-plangit clone --depth 1 https://github.com/tensormux/kernel-skillsWrote 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/tensormux/kernel-skills/write-backend-agnostic-kernel-plan)<a href="https://agentmods.dev/skills/tensormux/kernel-skills/write-backend-agnostic-kernel-plan"><img src="https://agentmods.dev/badge/skills/tensormux/kernel-skills/write-backend-agnostic-kernel-plan.svg" alt="Measured on agentmods" height="20"></a>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 | $0.00000 | $0.03004 |
| Opus 5 | $0.00000 | $0.01502 |
| Sonnet 5 | $0.00000 | $0.00601 |
| Haiku 4.5 | $0.00000 | $0.00300 |
Grade A, and why
write-backend-agnostic-kernel-plan 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 8d 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 — 151 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Skill: Write a Backend-Agnostic Kernel Plan
Purpose
Guide the agent through planning a compute kernel that must run correctly and performantly on multiple hardware backends (NVIDIA, AMD, CPU fallback, or future backends) before any backend-specific implementation is written — covering abstraction strategy, feature compatibility mapping, and the tradeoffs between portability and performance.
Use this when
- Designing a new kernel that must ship on both NVIDIA (CUDA) and AMD (ROCm/HIP) hardware.
- Evaluating whether Triton, OpenCL, SYCL, or manual multi-backend code is the right abstraction level for a given operation.
- The team needs a plan for supporting a new hardware backend without rewriting all existing custom kernels from scratch.
- Building a library or framework component that should not be tied to a single vendor's programming model.
Do not use this when
- NVIDIA is the only target and will remain so for the foreseeable future. Portable abstractions add engineering cost for no benefit.
- The operation can be expressed entirely through a framework like PyTorch (via ATen/Inductor) or JAX — backend portability is the framework's responsibility, not the kernel developer's.
- The performance requirement is so tight that the portability cost (abstraction overhead, suboptimal tile shapes per backend) cannot be absorbed.
Inputs the agent should gather first
- Required backends: which hardware targets must be supported? NVIDIA (which SMs), AMD (CDNA, RDNA), CPU, Apple Silicon, Intel GPU? Each pair of backends adds complexity.
- Operation type: elementwise, reduction, GEMM-like, attention, or custom. This determines how well it maps to existing cross-backend libraries.
- Performance requirements: is near-peak compute utilization required on all backends, or is correctness + reasonable performance sufficient? Maximum portability and maximum performance are inversely correlated.
- Team capabilities: does the team have expertise in all target backends, or only in one? A plan requiring intimate knowledge of AMD LDS banking and NVIDIA tensor core scheduling simultaneously is unrealistic for a small team.
- Tolerance for abstraction layers: is Triton acceptable? Is CUTLASS Cute? Is a fallback using cuBLAS/rocBLAS + a custom elementwise layer acceptable?
What ships with it
1 file 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.
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.
- 8d ago First seen · 151 lines · 0 tokens per session scan A abba92db0598
write-backend-agnostic-kernel-plan is a skill published in the GitHub repository tensormux/kernel-skills (73 stars, last pushed 2mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 3,004 tokens. 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
embedded-iot
Embedded systems firmware, microcontrollers (ESP32, STM32, Arduino, Raspberry Pi), RTOS (FreeRTOS, Zephyr), IoT protocols (MQTT, CoAP, BLE), bare-metal C/C++, and hardware peripheral interfaces (I2C, SPI, UART, GPIO). Use when developing firmware, working with microcontrollers, or building IoT devices.
hip-kernel-optimization
This skill should be used when writing or tuning HIP kernels on AMD/NVIDIA GPUs, covering memory coalescing, shared-memory tiling, bank conflict avoidance, warp primitives, occupancy, vectorization, async ops, loop unrolling, and profiling.
auditing-uefi-firmware-with-chipsec
Use Intel CHIPSEC to assess platform firmware configuration, SPI flash write protection, BIOS lock, SMM/SMRR, and Secure Boot variable state, dump SPI flash, and triage UEFI variables for firmware-level threats.
azure-local
Expert knowledge for Azure Local development including troubleshooting, best practices, decision making, architecture & design patterns, limits & quotas, security, configuration, integrations & coding patterns, and deployment. Use when planning Azure Local racks/SDN, disconnected clusters, Arc VMs, GPU workloads, or…
mi300-cdna3-architecture
MI300/CDNA3 architecture guide for HIP/Triton optimization—MFMA variants, dual register files, data formats, sparsity, LDS/GWS, and best practices.
mi300-hip-programming-insights
CDNA3/MI300 HIP programming insights—chiplet/cache model, Infinity Cache, memory coherency, matrix cores, sparsity, and best practices.