How do I install the Tensor Shape skill?

Run npx skillmds@latest add tenstorrent/tensor-shape in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Tensor Shape safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Tensor Shape?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Tensor Shape free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Tensor Shape?

tenstorrent (@tenstorrent) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Tensor Shape by tenstorrent — Agent Skill

Name: Tensor Shape
Author: tenstorrent

Tensor Shape

Convert LLK lib/API tile-size args to ckernel::TensorShape and maintain TRISC TensorShape coverage. Use when adding TensorShape parameters, replacing face_r_dim/num_faces, editing LLK_VALIDATE_TENSOR_SHAPE_*, regenerating tensor_shape_coverage_*.h, or reviewing TensorShape PRs.

tenstorrent 86dc35f 2.6 KB Updated Sep 17, 2026

File contents

/tensor-shape — TensorShape Plumbing & Coverage

Usage

/tensor-shape convert unpack_A
/tensor-shape coverage harvest unpack_A
/tensor-shape review

Mandatory First Step

Read the canonical reference and follow it:

Read(".claude/references/tensor-shape.md")

For metal-layer propagation after signature changes, also read .claude/references/metal-integration.md.

Modes

convert

Plumb TensorShape through an LLK op (lib → metal API → tests/fuser):

Replace face_r_dim / num_faces (and similar) with ckernel::TensorShape.
Include tensor_shape.h + the matching tensor_shape_coverage_{unpack|math|pack}.h.
Add LLK_VALIDATE_TENSOR_SHAPE_{UNPACK|MATH|PACK}("_llk_foo_...", tensor_shape) — string literal name; no API registry edit in coverage.h.
Prefer num_faces_r_dim / num_faces_c_dim for new logic; keep flat total_num_faces() only when intentionally preserving old mop behavior.
Update WH + BH together; Quasar only when in scope.
Metal wrappers: get_operand_tensor_shape(operand_id).
Tests/fuser: make_tensor_shape / cpp_tensor_shape; never legacy helper for 32×16.

coverage

Discover or regenerate TRISC allowlists:

Do not add a per-API enum entry.
Run discovery with TT_LLK_DISABLE_ASSERTS=1 + DPRINT, then: helpers/tensor_shape_coverage_parser.py harvest|emit|summary
New shape → named TENSOR_SHAPE_FR*_NF* in tensor_shape_coverage.h if needed + TRISC predicate update via emit.
New harvestable fn_name → add to MATH_FUNCTIONS / UNPACK_FUNCTIONS in the parser only.

review

Apply the Review Checklist in .claude/references/tensor-shape.md. Especially correct the misconception that every new API must be listed in coverage.h — coverage is TRISC-scoped; call sites only need the validate macro + string literal.

Key Files

Role	Path
Struct / helpers	`common/tensor_shape.h`
Macros / named shapes	`common/tensor_shape_coverage.h`
TRISC tables	`common/tensor_shape_coverage_{math,unpack,pack}.h`
Parser	`tests/python_tests/helpers/tensor_shape_coverage_parser.py`
Python TileShape	`tests/python_tests/helpers/tile_shape.py`
Metal operand helper	`tt_metal/hw/ckernels/{arch}/metal/llk_io/llk_operands.h`

---name: tensor-shapedescription: Convert LLK lib/API tile-size args to ckernel::TensorShape and maintain TRISC TensorShape coverage. Use when adding TensorShape parameters, replacing face_r_dim/num_faces, editing LLK_VALIDATE_TENSOR_SHAPE_*, regenerating tensor_shape_coverage_*.h, or reviewing TensorShape PRs.---# /tensor-shape — TensorShape Plumbing & Coverage## Usage```/tensor-shape convert unpack_A/tensor-shape coverage harvest unpack_A/tensor-shape review```## Mandatory First StepRead the canonical reference and follow it:```Read(".claude/references/tensor-shape.md")```For metal-layer propagation after signature changes, also read `.claude/references/metal-integration.md`.## Modes### convertPlumb `TensorShape` through an LLK op (lib → metal API → tests/fuser):1. Replace `face_r_dim` / `num_faces` (and similar) with `ckernel::TensorShape`.2. Include `tensor_shape.h` + the matching `tensor_shape_coverage_{unpack|math|pack}.h`.3. Add `LLK_VALIDATE_TENSOR_SHAPE_{UNPACK|MATH|PACK}("_llk_foo_...", tensor_shape)` — **string literal** name; **no** API registry edit in `coverage.h`.4. Prefer `num_faces_r_dim` / `num_faces_c_dim` for new logic; keep flat `total_num_faces()` only when intentionally preserving old mop behavior.5. Update WH + BH together; Quasar only when in scope.6. Metal wrappers: `get_operand_tensor_shape(operand_id)`.7. Tests/fuser: `make_tensor_shape` / `cpp_tensor_shape`; never legacy helper for 32×16.### coverageDiscover or regenerate TRISC allowlists:1. Do **not** add a per-API enum entry.2. Run discovery with `TT_LLK_DISABLE_ASSERTS=1` + DPRINT, then: `helpers/tensor_shape_coverage_parser.py harvest|emit|summary`3. New shape → named `TENSOR_SHAPE_FR*_NF*` in `tensor_shape_coverage.h` if needed + TRISC predicate update via emit.4. New harvestable `fn_name` → add to `MATH_FUNCTIONS` / `UNPACK_FUNCTIONS` in the parser only.### reviewApply the **Review Checklist** in `.claude/references/tensor-shape.md`. Especially correct the misconception that every new API must be listed in `coverage.h` — coverage is TRISC-scoped; call sites only need the validate macro + string literal.## Key Files| Role | Path ||------|------|| Struct / helpers | `common/tensor_shape.h` || Macros / named shapes | `common/tensor_shape_coverage.h` || TRISC tables | `common/tensor_shape_coverage_{math,unpack,pack}.h` || Parser | `tests/python_tests/helpers/tensor_shape_coverage_parser.py` || Python TileShape | `tests/python_tests/helpers/tile_shape.py` || Metal operand helper | `tt_metal/hw/ckernels/{arch}/metal/llk_io/llk_operands.h` |

Frequently asked questions