In my experience, a skill is better suited for this instead of an MCP.
If you don’t want the agent to probe the CLI when it needs it, a skill can describe the commands, arguments and flags so the agent can use them as needed.
For Jira/Confluence, I also struggled with their MCPs. JIRA’s MCPs was hit or miss and Confluence never worked for me.
We don’t use the cloud versions, so not sure if they work better with cloud.
On the other hand, i found some unofficial CLIs for both and they work great.
I wrote a small skill just to give enough detail about how to format Epics, Stories, etc and then some guidance on formatting content and I can get the agent do anything i need with them.
I deal with a ton of different atlassian instances and the most infuriating thing to me about the mcp configuration is that atlassian really thinks you should only have one atlassian instance to auth against. Their mcp auth window takes you to a webpage where you can’t see which thing you are authoring against forcing you to paste the login page url into an incognito window. Pretty half baked implementation.
I noticed that it’s better for some things than others. It’s pretty bad at working with confluence it just eats tokens but if you outlay a roadmap you want created or updated in Jira it’s pretty good at that
It's not for fun. They are hijacking a trusted server (Zendesk) to smuggle phishing links past my spam filter. Since Zendesk blocked the text relay, their bot is now just spamming signups as a side effect of the failed exploit.
This is interesting. I wrote a “memory indexer” with the idea to provide a tool (cli) for my agent to “remember” past conversations we had in other session. A little bit in the spirit of your second tool I think
The thought behind my aichat tool was to directly leverage the session log files and avoid creating any special memory artifacts. I do create a rust/tantivy index for the fast full text search.
The aichat approach does require intentionally asking the agent to find specific earlier work: it doesn’t automatically have “awareness” of prior work. I think of this as the “unknown unknowns” problem. This is where creating explicit memory artifacts can be useful since we can pre-inject recent work-summaries into context. So I’m thinking about a lightweight hook based system to automatically create memory artifacts or work-logs of some sort.
I can solve the cube with the regular “easy” 3-layer approach, but I’d like to solve it faster.
The issue is that the techniques for fast solving require to learn many different patterns to get to the right solution fast.
I don’t know really how ppl that solve it fast accomplish getting to that level, but to me it would be amazing if i could just set the cube in know scrambled states that let me practice and memorize specific algorithms repeatedly until I learn them.
The problem is that I don’t know enough yet to distinguish which are those initial states, let alone setting the cube in that state, so something that could set it up for me to practice would be amazing
> I don’t know really how ppl that solve it fast accomplish getting to that level
Just like everything else in life, they do it really slow and with lots and lots and lots of errors at first, but (and this is where the magic happens) keep doing it, training hours a day or their entire week ends, for years.
Maybe this may help. What if we are not talking internal development teams but something different, like a commercial/public API?
In those cases you cannot affort or expect to have meetings with folks to explian and communicate, and you also can appreciate more the abuse (unintended or not) that tokens can have.
I particularly liked that OP mentioned about expiration, key rotation and more advanced features you can achieve with his proposal, like switching schemes
If you don’t want the agent to probe the CLI when it needs it, a skill can describe the commands, arguments and flags so the agent can use them as needed.