Be aware of the AG products in use, and their versions and wrapper framework
Products in use may be obvious from the context, if not it can be determined from the package that features are imported from:
- packages start @ag-grid- or ag-grid-
- packages start ag-charts-
- package is ag-studio or starts ag-studio-
{product} below refers to grid, charts or studio.
Versions can be determined from project's package.json or by reading the installed library's package.json inside node_modules
Framework is
,
,
if using those frameworks,
for Vanilla JS apps or for apps on any other framework (e.g. Svelte, Solid).
For package version updates, use the ag-update skill
This skill ships with a sibling skill, "ag-update". Delegate to it when asked to update grid, charts or studio packages.
Don't guess
Whenever writing code, you must have a clear source for the APIs you use, whether that's following existing patterns, instructions from the user or consulting our docs. If you're unsure, do research to ground your actions.
By default consult the docs
When writing code you will encounter two problems:
- Your training data contains many deprecated and removed APIs and package names and may not have newer APIs.
- APIs have non-obvious edge cases and interactions with other features.
The solution to both of these is to consult the documentation.
If it is clear exactly what API to use, eg you are following a detailed plan that names specific APIs, or there are other examples in the codebase to copy, you may write code directly.
Otherwise if there is uncertainty, check the docs. To find the correct docs URL for the version in use, load
references/{product}/documentation-index.md
and follow the instructions in that file.
Locate the feature you are working with in the docs and read surrounding paragraphs to get information on edge cases and interactions. If many docs pages seem potentially relevant, consider getting a sub-agent to read them all and extract information relevant to the task.
Load product-specific recommendations
references/{product}/recommendations.md
contains specific advice for each product including known LLM failure modes and key APIs that have changed between versions. Load it and take it into account when developing.