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。
create-agent【开发与工程】MUST be used whenever scaffolding a new Atlas agent project. Creates agent.yaml, README.md, and a starter eval/cases.yaml with the correct structure, tool definitions, and instructions. Do NOT manually write agent.yaml from scratch — this skill handles the full scaffold. Triggers: create agent, new agent, scaffold agent, agent project, agent template…
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Scaffold a new Atlas agent project named $ARGUMENTS using npx @cognite/cli@latest agents create.
<agent-name>/
agent.yaml # Agent definition — externalId, model, instructions, tools
README.md # Quick-start docs and tool reference
eval/
cases.yaml # Starter eval cases (single-turn, multi-turn, correctness + faithfulness scorers)
The agent name is $ARGUMENTS. Ask the user for the following (skip any they already provided):
| Field | CLI flag | Default |
|---|---|---|
| Display name | --display-name | same as agent name |
| Description | --description | — |
| Model | --model | azure/gpt-5.4-mini |
| Instructions (system prompt) | --instructions | see default below |
| Tools | (edit after creation) | none |
Default instructions when the user doesn't specify:
Help users explore and understand the data in this project.
Use the available tools to retrieve data before answering.
When the data is insufficient, say so rather than guessing.
npx @cognite/cli@latest agents createRun the command, including only flags where the user provided a value or a non-empty default applies:
npx @cognite/cli@latest agents create <agent-name> \
--display-name "<display-name>" \
--description "<description>" \
--model <model> \
--instructions "<instructions>" \
--no-prompt
--display-name if the user wants the default (same as agent name).--description if none was given.--instructions (use the default if the user didn't specify).--no-prompt to skip the CLI's interactive prompts.--instructions and edit agent.yaml manually after creation.If the user wants tools, edit <agent-name>/agent.yaml to add a tools array. Do NOT add tools: [] — omit the key entirely when there are no tools.
| Type | Purpose |
|---|---|
analyzeData | Analyze tabular or structured data |
analyzeImage | Analyze images and P&ID diagrams |
analyzeTimeSeries | Analyze time series data |
askDocument | Ask questions about documents |
callFunction | Call a Cognite Function |
callRestApi | Call an external REST API |
callWebhook | POST a payload to an external webhook |
examineDataSemantically | Semantic data examination |
query | Structured queries against CDF data models |
queryKnowledgeGraph | Query CDF data models with natural language |
queryTimeSeriesDatapoints | Fetch raw or aggregated time series data |
runPythonCode | Execute custom Python code |
summarizeDocument | Summarize documents |
timeSeriesAnalysis | Advanced time series analysis and anomaly detection |
tools:
- name: find_assets
type: queryKnowledgeGraph
description: Find assets and related instances in the knowledge graph.
configuration:
version: v2
dataModels:
- space: cdf_cdm
externalId: CogniteCore
version: v1
viewExternalIds:
- CogniteAsset
instanceSpaces:
type: all
agent.yaml, README.md, and eval/cases.yaml were all created<agent-name>/agent.yaml — add tools and refine instructionsnpx @cognite/cli@latest agents push <externalId> — push to CDF for testingnpx @cognite/cli@latest agents open — open in Fusion<agent-name>/eval/cases.yaml and run npx @cognite/cli@latest agents eval <externalId> — the scaffolded cases are placeholders (a correctness case, a faithfulness case, and a multi-turn case); update the inputs, references, and context to match what the agent actually doestools: [] — omit the key entirely when there are no tools