菜单
Create and adapt the inference snap structure from reference examples (packaging, engines, hooks, scripts, components).
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。
基于 SOC 职业分类
Create GitHub Actions workflow files for inference snap repositories based on canonical/gemma4-snap patterns.
Build, install, and runtime-verify an inference snap including engine selection and prompt/API checks.
Run the full inference-snap workflow as a sequential chain of subagents — structure → github workflows → static checks → build & test — passing each stage's report as input to the next.
Perform static quality checks for an inference snap (snapcraft.yaml consistency, component completeness, model signature verification).
| name | inference-snap-structure |
| description | Create and adapt the inference snap structure from reference examples (packaging, engines, hooks, scripts, components). |
| trigger | Keywords like "create snap structure", "adapt inference snap", "scaffold inference snap", "set up engines/hooks/components" |
| scope | user |
Create the snap structure by reusing proven patterns from example repositories, with minimal targeted adaptation.
RULESET.md (sibling file in this skill directory) — the authoritative structural ruleset extracted from the original example repos. Read it before scaffolding. It defines Variant A (single-file model) and Variant B (sharded model > 5 GB) and is self-contained — do NOT consult external example repos.RULESET.md and pick the variant (A: model ≤ 5 GB, B: model > 5 GB).dev submodule targeting https://github.com/canonical/inference-snaps-dev.git)git clone --recurse-submodules instructions){{SNAP_NAME}}, {{MODEL_ALIAS}}, {{PORT}}, …).dev git submodule in the new project.