协作头脑风暴
brainstorming
在实现前澄清需求、探索设计选择,并形成清晰的开发方向。
agent-workflows【开发与工程】Guide Wren CLI-served agent workflow guides, discovery-stub installation, wren skills retrieval, guided prompt shaping, dlt SaaS ingestion, and Wren skill authoring.
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Use this sub-skill when an agent needs Wren's own workflow guides, the wren skills command, the discovery stub, wren ask, dlt/SaaS ingestion, or authoring
new guide content that ships with the CLI.
Install the CLI first, then inspect the version-matched guide inventory:
pip install wrenai
wren skills list
wren skills get onboarding
wren skills get usage --full
wren skills get dlt-connector --script introspect_dlt
The current guide names are onboarding, usage, generate-mdl,
dlt-connector, enrich-context, and genbi. Use --full for bundled
references and --script to print a named bundled helper.
| Intent | Guide |
|---|---|
| First install, connection, project, and first query | onboarding |
| Day-to-day governed data question | usage |
| Build MDL from a database schema | generate-mdl |
| Load SaaS data through dlt and create a Wren project | dlt-connector |
| Add business rules, definitions, or named metrics | enrich-context |
| Build or deploy a dashboard application | genbi |
Read the returned guide before carrying out its multi-step workflow. These guides are package data, so their contents match the installed CLI version.
wren ask produces a prompt; it does not query a database:
wren ask "How many active customers did we have last month?" --guided
wren ask "How many active customers did we have last month?" --direct
Choose exactly one mode. --guided supplies a stricter workflow for a weaker
agent; --direct is a minimal wrapper for a stronger agent.
For SaaS data, use the dlt guide before assuming a direct Wren connector exists. The normal flow is source API -> dlt -> local DuckDB -> generated Wren project -> validate/build/query. Use the bundled helper only with a local DuckDB file and review generated models before a live query.
New Wren workflow guides ship as package data, not as discovery-stub directories.
Use the CLI skill-content location in the package source when maintaining the
repository, add a matching guide name, and test wren skills get delivery.
references/agent-skills.md for install and delivery behavior.references/wren-ask.md for explicit prompt-mode rules.references/dlt-workflow.md before creating a project from SaaS/DuckDB
data.references/skill-authoring.md when changing a CLI-served guide.references/troubleshooting.md for stub, retrieval, dlt, and credential
failures.scripts/introspect_dlt_project.py --help before using its local
DuckDB-to-project generator.wrenai; the
stub only teaches a compatible agent how to discover the CLI.wren command.