Templates

<!-- TEMPLATE: the entry-router skill (`<fw>`). This is the most important skill in the family -- it routes every request to the right specialist. Directory: skills/<fw>/SKILL.md The frontmatter `name` MUST equal the directory name (the bare prefix `<fw>`). Fill placeholders from capability-inventory.md, then DELETE this comment. -->

---

name: <fw> description: >- Starting point for ANY work with <Framework> -- <list the main capabilities / primitives>. Routes to the right specialist skill via a decision tree. Use proactively whenever the user mentions <Framework>, <its primitives>, <its file/config types>, or asks "which <primitive>", "how do I <core task>", or "should I use <A> or <B>". Read this first to pick the right approach and route to the granular <fw>-* skills.

<Framework>

<One-paragraph description of what the framework is and the value it provides.>

BEFORE YOU START

• Read this whole file before acting.
• Resolve all environment-specific values (paths, namespaces, identities) from the config object -- never hardcode them. See references/config-and-paths.md.
• For deep field-level detail you cannot find here, call the runtime doc API: <the framework's get_docs(...) / list_docs(...) call>.

Mental model

<3-6 lines or a small ASCII diagram: the core primitives and how data/work flows through the named stages. Name the primitive that owns each stage.>

Which approach? (decision tree)

What do you need to do?
├── <ingest from a source>?      -> <primitive-A>   (start: <fw>-onboarding)
├── <build a curated product>?   -> <primitive-B>   (start: <fw>-<product>)
├── <general pipeline pattern>?  -> <primitive-C>   (start: <fw>-<pipeline>)
├── <schedule / orchestrate>?    -> <primitive-D>   (start: <fw>-orchestration)
├── <add quality checks>?        -> <inline> or <library> (see <fw>-data-quality)
└── <explore before authoring>?  -> <fw>-explore-* skills

I want to ... (CUJ -> skill index)

I want to ...	Start here
Onboard a new source end to end	<fw>-onboarding
Author config/metadata	<fw>-metadata
Transform / cast columns	<fw>-expressions
Add data-quality checks	<fw>-data-quality
Size / tune compute	<fw>-compute
Schedule a job	<fw>-orchestration
Restrict row access	<fw>-security
Monitor / diagnose runs	<fw>-ops
Decommission a system	<fw>-offboarding
<add one row per shipped skill>	<fw>-________

Quick reference

<The single most common minimal example -- e.g. the smallest valid config or the canonical "hello world" call. Complete and copy-pasteable.>

Artifact layout

<Inline the real file/config layout as an ASCII tree so a source-blind agent knows where things go. Include per-entity vs per-system folders, canonical file names, and any naming/FQN rules.>

Environment & promotion

• Tiers: <e.g. dev -> test -> acceptance -> prod>. Promote in order; never production first.
• Per-env values resolve from the config object: <how>.

Guardrails

• <Destructive operations> require explicit user consent; always propose a non-destructive alternative first.
• Default to read-only exploration before any write.

Reference library

Guide	Read it when
references/config-and-paths.md	resolving env-specific paths/namespaces
references/migration.md	converting raw/ad-hoc code into <Framework> primitives
<add deep references as needed>

Related skills

<fw>-onboarding, <fw>-metadata, <fw>-data-quality, <fw>-compute, <fw>-orchestration, <fw>-security, <fw>-ops <... list the family>.