菜单
Scaffold a new T2 (L1-direct) Op file from a single `src/tileops/manifest/` entry by following the 7-step playbook in docs/design/ops-design.md. Emits the 17 scaffold slots (S1-S7, S12-S21); leaves family-specific protocol variables, optional hooks, and kernel implementations to downstream skills.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/tile-ai/TileOPs --skill scaffold-op命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
Generate or re-align one `src/tileops/manifest/` entry from a reference-API docs URL. Caller provides the manifest key (`op_name`); skill writes that one entry. Idempotent.
Write tests for the target spec using PyTorch as ground truth, verify they fail on current code.
Drive the full migration for an op family — audit, delegate per-op alignment to align-op, run cross-op cleanup. Two terminal outcomes: SUCCESS opens a PR; CLEANUP_REGRESSION exits blocked without a PR when post-cleanup tests fail.
基于 SOC 职业分类
| name | scaffold-op |
| description | Scaffold a new T2 (L1-direct) Op file from a single `src/tileops/manifest/` entry by following the 7-step playbook in docs/design/ops-design.md. Emits the 17 scaffold slots (S1-S7, S12-S21); leaves family-specific protocol variables, optional hooks, and kernel implementations to downstream skills. |
op_name (positional) — manifest key for the op to scaffold, equal to the target cls.__name__ (e.g. CumsumFwdOp).
op_name must be present in src/tileops/manifest/ with status: spec-only and a non-empty source.kernel_map. source.kernel_map is manifest-level source of truth for Op→Kernel dispatch and cannot be derived by the scaffold (dispatch keys are kernel-internal conventions), so a spec-only entry needs it added before the scaffold can run.src/ + manifest source.op (e.g., src/tileops/ops/reduction/cumulative.py), containing the 17 scaffold slots; one-line from .<module> import <ClassName> added to the package __init__.py at that file's parent directory (e.g., src/tileops/ops/reduction/__init__.py) with a matching __all__ entry. Note: that parent directory is not always the same as the manifest family field — for example, CumsumFwdOp has family: scan but lives under src/tileops/ops/reduction/. Always key paths off source.op, never off family. Plus a side-artefact at .foundry/plan/<op_name>/plan.json carrying the DRY_RUN self-audit (not tracked in git).python scripts/validate_manifest.py --check-op <op_name> reports no errors for this op. Warnings are allowed and passed through to the final summary.op_name that the scaffold cannot fix by re-reading the playbook's slot rules. Do NOT commit; report with the failing rows from the validator._op_kind, _kernel_key, _kernel_cls, _kernel_handles_padding, _op_name, kernel_cls)._pad_value, _validate_dim, _pre_kernel, _post_kernel, _cache_key override).src/tileops/manifest/, tests, benchmarks, or any existing op file.stateDiagram-v2
[*] --> READ
READ --> PRE_CHECK
PRE_CHECK --> DRY_RUN
PRE_CHECK --> BLOCKED: precondition failed
DRY_RUN --> EMIT: plan.json written
EMIT --> REGISTER
REGISTER --> VALIDATE
VALIDATE --> REPORT
VALIDATE --> BLOCKED: validator error or §1 drift
REPORT --> [*]
BLOCKED --> [*]
DRY_RUN writes plan.json to freeze manifest-sourced facts before codegen. VALIDATE diffs the emitted file against plan.json §1 — any drift is a skill bug, not a manifest issue.
Emit exactly the 17 slots in slot-rules.md: S1–S7, S12–S21. S8–S11 are reserved for T1 thin-wrapper subclasses and skipped.
Out of scope — leave empty:
| Item | Reason |
|---|---|
Family protocol vars (_op_kind, _kernel_key, _op_name, …) | Kernel-dispatch convention; not in manifest |
Optional hooks (_pad_value, _validate_dim, _pre_kernel, _post_kernel) | Op-specific business logic |
_cache_key override | Recommended under dynamic shapes; depends on kernel math |
| Kernel implementations | Owned by kernel skill |
| Tests / benchmarks | Owned by test-op / bench-op |
These gaps surface as NotImplementedError or validator warnings; downstream skills fill them.
Load the manifest entry for op_name:
Before running the snippet, substitute <op_name> with the requested manifest key (the skill's positional argument — agent literal substitution, not shell interpolation):
python - "<op_name>" <<'PY'
import sys
from tileops.manifest import load_manifest
op_name = sys.argv[1]
entry = load_manifest()[op_name]
print(entry)
PY
Extract: family, status, signature.inputs, signature.outputs, signature.params, signature.static_dims, signature.shape_rules, source.kernel_map, source.op, source.kernel, roofline.vars, roofline.flops, roofline.bytes.
Derive the target file path by prepending src/ to source.op (e.g. src/tileops/ops/reduction/cumulative.py). The filesystem package directory is that file's parent (e.g. src/tileops/ops/reduction/). Do not use the manifest family field to compute paths — it is a semantic label, and some ops have family distinct from their filesystem parent (e.g., CumsumFwdOp has family: scan but lives under reduction/). Module filename is source.op's basename without .py.
op_name present in src/tileops/manifest/ → proceed; otherwise BLOCKED ("op not in manifest").status field explicitly set to spec-only → proceed; status: implemented → BLOCKED ("op already implemented; use implement-op to migrate"); missing status or any other value → BLOCKED ("manifest entry must declare a valid top-level status; the validator treats status as required").source.kernel_map declared and non-empty → proceed; missing or empty → BLOCKED ("manifest entry needs source.kernel_map before scaffolding — add the dispatch map to the manifest first; the scaffold cannot invent dispatch keys because they are kernel-internal conventions"). Note: per docs/design/manifest.md, source.kernel_map is only required when status: implemented, so many existing spec-only entries lack it — these are the cases that need the map added before scaffolding can run.source.kernel_map resolves to an importable symbol → proceed; otherwise BLOCKED ("kernel class not found at expected path").source.op does NOT exist → proceed; exists → BLOCKED ("target file already present; scaffold would overwrite").BLOCKED terminations return without writing any file.
Write .foundry/plan/<op_name>/plan.json with three sections:
| Section | Diffed at VALIDATE? | Content |
|---|---|---|
locked_facts (§1) | yes (hard error on drift) | Verbatim manifest extraction (op_name, class_name, family, module_path, kernel_imports, kernel_map, init_kwargs, forward_inputs/outputs, dtype_unions, dtype_combos, shape_rules, static_dims, roofline) |
agent_notes (§2) | no | Judgement calls (docstring, kernel ctor signature observed, helper state, forward reshape strategy, codebase refs consulted) |
open_questions (§3) | no | Ambiguities tagged needs_doc_fix / needs_manifest_fix / needs_human_decision; surfaced in REPORT, never block |
Always proceed to EMIT. Empty open_questions is fine.
Skeleton:
{
"locked_facts": {
"op_name": "CumsumFwdOp",
"module_path": "src/tileops/ops/reduction/cumulative.py",
"kernel_map": {"cumulative_fwd": "CumulativeKernel"},
"init_kwargs": [{"name": "dim", "source": "signature.params.dim", "type": "int", "default": -1}],
"forward_inputs": ["x"],
"static_dims": {"N": "x.shape[dim]"},
"roofline": {"flops": "M * N", "bytes": "2 * M * N * elem_bytes"}
},
"agent_notes": {"docstring_summary": "...", "kernel_ctor_signature_observed": "..."},
"open_questions": [{"tag": "needs_human_decision", "topic": "...", "detail": "..."}]
}
Follow docs/design/ops-design.md § Scaffolding an Op from a Manifest Entry Steps 1-7 in order. For each scaffold slot, read the authoritative rule at slot-rules.md#slot-sN before emitting.
Key slot pointers (follow the reference, do not re-derive):
| Playbook step | Slots | Reference anchor |
|---|---|---|
| Step 1 | S1, S2, S3, S4 | S1-S4 |
| Step 2 | S5, S6, S7 | S5-S7 |
| Step 3 | S21, S12, S13 | S21, S12, S13 |
| Step 4 | S14, S15, S16 | S14-S16 |
| Step 5 | S17, S18 | S17, S18 |
| Step 6 | S19 | S19 |
| Step 7 | S20 | S20 |
If a slot's rule is ambiguous for the given manifest entry (e.g. multi-kernel kernel_map, multiple independent dtype axes, fixed-rank vs arbitrary-rank branching), STOP and surface the ambiguity in the final report instead of guessing. Do not expand scope.
Append to the __init__.py at dirname(source.op):
| File style | Action |
|---|---|
Grouping comments # --- <KernelClassName> ops --- (e.g. src/tileops/ops/reduction/__init__.py) | Place import under matching kernel block; create a new block (alphabetical) when no existing block references the kernel |
Flat imports (e.g. src/tileops/ops/norm/__init__.py) | Append alongside existing imports following file's ordering convention; do NOT introduce grouping comments |
Add matching <ClassName> to __all__, preserving any existing sectioning.
(a) §1 post-check — diff plan.json.locked_facts against the emitted artefacts:
source.op with ast: extract class name, base, imports, __all__, __init__ kwargs (names / defaults / types in order), forward params, default_kernel_map dict, _static_axes.dirname(source.op)/__init__.py with both ast (presence + __all__ membership) and raw text (block placement under # --- <KernelClassName> ops --- when the file uses grouping comments; skip placement check on flat-style files).locked_facts field mismatch → BLOCKED §1 drift: <field>. Skill deviated from its own contract; fix the emitted file (or revert and restart DRY_RUN if the plan was wrong). Do NOT edit plan.json to match.(b) Manifest validator:
python scripts/validate_manifest.py --check-op <op_name>
--check-op runs L0–L4 even on status: spec-only. Classify output:
_infer_output_shapes / _validate_dtypes against manifest shape_rules / dtype_combos first — likely a mis-emit.Never edit the manifest to silence an error. No parity_opt_out escape hatch. Demote to status: spec-only only when the implementation genuinely cannot conform.
Status: SUCCESS | BLOCKED
Op: <op_name>
File: <path> (<lines>)
Package registration: <parent-of-source.op>/__init__.py (+1 import, +1 __all__)
Plan: .foundry/plan/<op_name>/plan.json
§1 drift: <none | <field>: <plan> vs <emitted>>
Validator: <N errors, M warnings>
<if BLOCKED, list each blocking error row verbatim>
Warnings:
- <warning>
Open questions (from plan.json §3):
needs_doc_fix: - <topic>: <detail>
needs_manifest_fix: - <topic>: <detail>
needs_human_decision: - <topic>: <detail>
Not filled (downstream hand-off):
- Kernel implementation
- Optional hooks (family-specific)
- Family protocol variables (family-specific)
- _cache_key override (recommended under dynamic shapes)
SUCCESS → commit [Feat][OPS] scaffold <op_name>. BLOCKED → leave tree dirty, return. plan.json persists either way.
status: implemented ops use implement-op / align-family.If a slot rule is internally inconsistent or cannot be followed mechanically, file a doc issue and BLOCK; do not invent slot behaviour.