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 agentmods add skills/forceinjection/cuda-code-skill/cuda-knowledgenpx skills add ForceInjection/cuda-code-skill --skill cuda-knowledgegit clone --depth 1 https://github.com/ForceInjection/cuda-code-skillWhat 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 | $0.00249 | $0.03874 |
| Opus 5 | $0.00125 | $0.01937 |
| Sonnet 5 | $0.00050 | $0.00775 |
| Haiku 4.5 | $0.00025 | $0.00387 |
Grade A, and why
cuda-knowledge 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 yesterday.
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.
The source is not reproduced here
No licence file
A repository with no LICENSE is all rights reserved by default, so the body is not copied here. The metadata, the measurements and the link are.
What ships with it
60 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.
- references/cublas-docs/1-introduction/1-introduction.md 24 KB
- references/cublas-docs/1-introduction/1.1-data-layout.md 1.1 KB
- references/cublas-docs/1-introduction/1.2-new-and-legacy-cublas-api.md 3.4 KB
- references/cublas-docs/1-introduction/1.3-example-code.md 5.2 KB
- references/cublas-docs/1-introduction/1.4-forward-compatibility.md 377 B
- references/cublas-docs/1-introduction/1.5-floating-point-emulation.md 12 KB
- references/cublas-docs/1-introduction/1.5.1-bf16x9.md 1.3 KB
- references/cublas-docs/1-introduction/1.5.2-fixed-point.md 5.6 KB
- references/cublas-docs/1-introduction/1.5.2.1-dynamic-mantissa-control.md 630 B
- references/cublas-docs/1-introduction/1.5.2.2-fixed-mantissa-control.md 457 B
- references/cublas-docs/1-introduction/1.5.2.3-representation-and-mappings.md 644 B
- references/cublas-docs/1-introduction/1.5.2.4-fixed-point-workspace-requirements.md 565 B
- references/cublas-docs/1-introduction/1.5.2.5-fixed-point-performance-guide.md 2.4 KB
- references/cublas-docs/1-introduction/1.5.3-default-library-configurations.md 1.2 KB
- references/cublas-docs/1-introduction/1.5.4-support-for-floating-point-special-values.md 897 B
- references/cublas-docs/2-using-the-cublas-api/2-using-the-cublas-api.md 428 KB
- references/cublas-docs/2-using-the-cublas-api/2.1-general-description.md 19 KB
- references/cublas-docs/2-using-the-cublas-api/2.1.1-error-status.md 122 B
- references/cublas-docs/2-using-the-cublas-api/2.1.10-gemm-algorithms-numerical-behavior.md 1.1 KB
- references/cublas-docs/2-using-the-cublas-api/2.1.11-tensor-core-usage.md 1.6 KB
- references/cublas-docs/2-using-the-cublas-api/2.1.12-cuda-graphs-support.md 1.6 KB
- references/cublas-docs/2-using-the-cublas-api/2.1.13-bit-integer-interface.md 1.8 KB
- references/cublas-docs/2-using-the-cublas-api/2.1.2-cublas-context.md 2.1 KB
- references/cublas-docs/2-using-the-cublas-api/2.1.3-thread-safety.md 655 B
- references/cublas-docs/2-using-the-cublas-api/2.1.4-results-reproducibility.md 2.4 KB
- references/cublas-docs/2-using-the-cublas-api/2.1.5-scalar-parameters.md 2.2 KB
- references/cublas-docs/2-using-the-cublas-api/2.1.6-parallelism-with-streams.md 1.3 KB
- references/cublas-docs/2-using-the-cublas-api/2.1.7-batching-kernels.md 1.7 KB
- references/cublas-docs/2-using-the-cublas-api/2.1.8-cache-configuration.md 786 B
- references/cublas-docs/2-using-the-cublas-api/2.1.9-static-library-support.md 1.5 KB
- references/cublas-docs/2-using-the-cublas-api/2.2-cublas-datatypes-reference.md 15 KB
- references/cublas-docs/2-using-the-cublas-api/2.2.1-cublashandle_t.md 401 B
- references/cublas-docs/2-using-the-cublas-api/2.2.10-cublasmath_t.md 2.2 KB
- references/cublas-docs/2-using-the-cublas-api/2.2.11-cublascomputetype_t.md 3.3 KB
- references/cublas-docs/2-using-the-cublas-api/2.2.12-cublasemulationstrategy_t.md 1.1 KB
- references/cublas-docs/2-using-the-cublas-api/2.2.2-cublasstatus_t.md 2.6 KB
- references/cublas-docs/2-using-the-cublas-api/2.2.3-cublasoperation_t.md 595 B
- references/cublas-docs/2-using-the-cublas-api/2.2.4-cublasfillmode_t.md 533 B
- references/cublas-docs/2-using-the-cublas-api/2.2.5-cublasdiagtype_t.md 513 B
- references/cublas-docs/2-using-the-cublas-api/2.2.6-cublassidemode_t.md 509 B
- references/cublas-docs/2-using-the-cublas-api/2.2.7-cublaspointermode_t.md 687 B
- references/cublas-docs/2-using-the-cublas-api/2.2.8-cublasatomicsmode_t.md 478 B
- references/cublas-docs/2-using-the-cublas-api/2.2.9-cublasgemmalgo_t.md 1.8 KB
- references/cublas-docs/2-using-the-cublas-api/2.3-cuda-datatypes-reference.md 4.7 KB
- references/cublas-docs/2-using-the-cublas-api/2.3.1-cudadatatype_t.md 1.9 KB
- references/cublas-docs/2-using-the-cublas-api/2.3.2-cudaemulationstrategy_t.md 228 B
- references/cublas-docs/2-using-the-cublas-api/2.3.3-cudaemulationmantissacontrol_t.md 751 B
- references/cublas-docs/2-using-the-cublas-api/2.3.4-cudaemulationspecialvaluessupport_t.md 1.3 KB
- references/cublas-docs/2-using-the-cublas-api/2.3.5-librarypropertytype_t.md 388 B
- references/cublas-docs/2-using-the-cublas-api/2.4-cublas-helper-function-reference.md 31 KB
- references/cublas-docs/2-using-the-cublas-api/2.4.1-cublascreate.md 1.4 KB
- references/cublas-docs/2-using-the-cublas-api/2.4.10-cublasgetpointermode.md 528 B
- references/cublas-docs/2-using-the-cublas-api/2.4.11-cublassetpointermode.md 648 B
- references/cublas-docs/2-using-the-cublas-api/2.4.12-cublassetvector.md 1.1 KB
- references/cublas-docs/2-using-the-cublas-api/2.4.13-cublasgetvector.md 1.1 KB
- references/cublas-docs/2-using-the-cublas-api/2.4.14-cublassetmatrix.md 1.0 KB
- references/cublas-docs/2-using-the-cublas-api/2.4.15-cublasgetmatrix.md 1.0 KB
- references/cublas-docs/2-using-the-cublas-api/2.4.16-cublassetvectorasync.md 784 B
- references/cublas-docs/2-using-the-cublas-api/2.4.17-cublasgetvectorasync.md 784 B
- references/cublas-docs/2-using-the-cublas-api/2.4.18-cublassetmatrixasync.md 810 B
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.
- yesterday First seen · 311 lines · 249 tokens per session scan A f0d76ba00a88
cuda-knowledge is a skill published in the GitHub repository ForceInjection/cuda-code-skill (24 stars, last pushed 1mo ago), with no licence file. It adds 249 tokens to every session and 3,874 once invoked, about $0.0012 per session on Opus 5. A static security scan graded it A with 0 findings. It comes from a forked repository.
Other skills, from other repositories
kernel-KBS
Corpus-backed GPU kernel knowledge base for CUDA, Triton, CuTe, CUTLASS, and Ampere/Hopper/Blackwell kernel research. Use when the user needs to search merged kernel PR pages, inspect PR diff/provenance artifacts, find KernelWiki synthesis pages, query blog/doc/contest notes, or retrieve evidence-backed implementation…
kernel-benchmark
Standalone kernel benchmarking skill for cuda-cpp, cutlass, cute-dsl, and triton implementations. Use when the user wants to compare a custom CUDA/CUTLASS .cu kernel or CuTe DSL/Triton .py kernel against selectable PyTorch eager, torch.compile, or FlashInfer baselines, validate correctness, measure execution time with…
kernel-loop
Iterative GPU kernel optimization orchestrator for CUDA/CUTLASS/CuTe DSL/Triton kernels. Use for measured, one-change-at-a-time optimization loops with correctness, NCU profiling, KBS evidence, hypothesis discipline, hard iteration gates, final benchmarking, and a traceable report.
kernel-profile
Standalone kernel profiling skill for cuda-cpp, cute-dsl, cutlass, and triton implementations. Checks CUDA/PyTorch/Triton/CuTe DSL/CUTLASS/NCU/nsight-python readiness, optionally locks GPU clocks, validates correctness, collects Nsight Compute metrics with nsight-python, produces envcheck.md, correctness.md…
mlir-opt-skill
Use when optimizing MLIR GPU performance, compiling PyTorch or linalg/tensor MLIR to PTX, writing Transform dialect schedules, tuning auxiliary pass pipelines, benchmarking MLIR kernels against PyTorch, or diagnosing linalg/torch-mlir GPU codegen.
claude-md-improver
Audit and improve CLAUDE.md files in repositories. Use when user asks to check, audit, update, improve, or fix CLAUDE.md files. Scans for all CLAUDE.md files, evaluates quality against templates, outputs quality report, then makes targeted updates. Also use when the user mentions "CLAUDE.md maintenance" or "project…