Include best practices for agentic coding assistants#381
Conversation
Added guidelines for using agentic coding assistants on shared supercomputer systems, emphasizing responsible usage and best practices to avoid disruption.
|
First take on advice for using agentic coding assistants on CISL resources (Derecho / Casper). Long enough that it might feasibly be in it's own section, particularly as we expand agentic content, but stuck it here for now. |
|
Some initial comments as I have not yet dived into the content yet. This "Best Practices" page is likely a little hard to find for most users. The page itself was already a little long. Instead, what do you thing of making a new sidebar head section "Best Practices" then having sub-pages for each of the sections under that only 1 level deep? This would at least make the LLM recommendation page more visible to users from the sidebar. |
|
Implementing the page with a sidebar is a good idea! I hope it will also
allow searches to find the page, under *agentics" or "LLM" etc.
…On Wed, Jul 1, 2026 at 4:23 PM Daniel Howard ***@***.***> wrote:
*dphow* left a comment (NCAR/HPC-Docs#381)
<#381 (comment)>
Some initial comments as I have not yet dived into the content yet.
This "Best Practices" page is likely a little hard to find for most users.
The page itself was already a little long. Instead, what do you thing of
making a new sidebar head section "Best Practices" then having sub-pages
for each of the sections under that only 1 level deep? This would at least
make the LLM recommendation page more visible to users from the sidebar.
—
Reply to this email directly, view it on GitHub
<#381?email_source=notifications&email_token=AEHSBB5ZQCTMK4ASFBT4AR35CWFOBA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIOBWGA2TAOBRGU2KM4TFMFZW63VKON2WE43DOJUWEZLEUVSXMZLOOSWGM33PORSXEX3DNRUWG2Y#issuecomment-4860508154>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AEHSBB4MPNGF2D6RDII57AT5CWFOBAVCNFSNUABFKJSXA33TNF2G64TZHM2TQOBTGYYTSOJQHNEXG43VMU5TINZVGUYTSNRXG422C5QC>
.
Triage notifications, keep track of coding agent tasks and review pull
requests on the go with GitHub Mobile for iOS
<https://github.com/notifications/mobile/ios/AEHSBB2HRZ4O55A7JHYYSQT5CWFOBA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIOBWGA2TAOBRGU2KM4TFMFZW63VKON2WE43DOJUWEZLEUVSXMZLOOSVGM33PORSXEX3JN5ZQ>
and Android
<https://github.com/notifications/mobile/android/AEHSBB3ZK4VGFZJ7H4YDHXT5CWFOBA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIOBWGA2TAOBRGU2KM4TFMFZW63VKON2WE43DOJUWEZLEUVSXMZLOOSXGM33PORSXEX3BNZSHE33JMQ>.
Download it today!
You are receiving this because you are subscribed to this thread.Message
ID: ***@***.***>
|
|
I committed additional edits re-organizing the nav bar and some other edits across pages while doing this. On the content piece, I mainly edited some of the sections content in minor ways so as to remove emdashes and make it look less AI assisted in the writing sense. Otherwise, the text makes sense and is now easily findable from main pages. I think the current pages Rory made cover similar points from https://docs.ycrc.yale.edu/ai/aicodingtools/ and https://scicomp.aalto.fi/triton/usage/ai-agents/. If other folks review those pages from peers and think we should explicitly include similar content, happy to hear that feedback. |
Would it be a good idea to tell it to delete secrets from the environment when the task is finnished?
Added guidelines for using agentic coding assistants on shared supercomputer systems, emphasizing responsible usage and best practices to avoid disruption.