Discussion about this post

User's avatar
Paul Sturrock's avatar

I’ve been following this series and it’s been extremely helpful.

I started setting up something very similar with Cowork’s productivity plug in. But I ran into trouble because it kept getting confused as to directories. It seemed to be because I was using iCloud Drive. I was doing so not only for back up reasons, but so that I could access info on my phone (via Obsidian).

I switched to Claude Chat, and it’s been working great. I upload my ClaudeMD file, daily note, and if relevant projectmd files. We chat about the priorities etc. and Claude gives me updated files with briefings for the day, week etc.

The advantage of chat seems to be that it uses session and token limits more slowly. The disadvantage seems to be that it is a slight faff to upload and download the files.

Did you run into similar issues with Cowork finding directories? How do you decide when to choose between Code, Cowork and Chat these days?

Pawel Jozefiak's avatar

The skills system is essentially CLAUDE.md but structured for human-readable reuse rather than session injection. I've been running a version of this for months.

What I found interesting is that the underlying model matters more than you'd expect for skill execution. Same skill file, different model - meaningfully different quality of autonomous decision-making. With Sonnet 4.6, I noticed the morning routine skill started making judgment calls I hadn't explicitly coded for. Not wrong calls, just more autonomous ones.

Your three-tier structure (morning, evening, weekly) is solid. One thing worth adding: exception handling. What does the agent do when the daily pattern breaks? That's where most skill systems quietly fail.

Related experiment from this week: https://thoughts.jock.pl/p/sonnet-46-two-experiments-one-got-personal

6 more comments...

No posts

Ready for more?