| name | tensor4all-rules |
| description | Use when working in Tensor4all repositories or related Rust/Julia tensor projects and you need current shared agent rules for repository workflow, performance, numerical correctness, docs, tests, benchmarks, GPU kernels, or language bindings. Fetch/read the latest rules first, and load only the rule files relevant to the task. |
Tensor4all Rules
Use this skill as a thin trigger for the shared Tensor4all agent rules. The
rules live outside the skill body so they can be updated and loaded
progressively.
Load Rules
- Prefer the latest online
tensor4all-agent-rules repository when internet
access is available.
- If internet access is unavailable, look for a sibling checkout at
../tensor4all-agent-rules.
- Read
rules/index.md first.
- Load only the common, Rust, Julia, performance, numerical, docs, or benchmark
rule files needed for the current task.
Apply Rules
- Treat project-local
AGENTS.md, REPOSITORY_RULES.md, and equivalent files
as more specific overrides.
- Do not bulk-load all rule files unless the task is explicitly to audit or
update the shared rules.
- When a rule conflicts with local project instructions, follow the local
instruction and note the conflict if it affects the result.