adeleinr/coding_agent_setup.txt

## coding_agent_setup.txt
"There is no playbook for building AI native companies, and if there is, I dont have it" ~ Dan Shipper, https://www.youtube.com/watch?v=MGzymaYBiss

But, here are some good practices learned so far:

1) One approacj is to use a single Claude.md in the repo.
   Note patterns you want AND patterns you do not want Claude to use as you develop in it.
   Should Contain setup necessary to run the repo, to be used by new engineers (Reduces barried of entry)
   Every member of the team contributes to it from lessons learned.
2) Tagging @.claude on a PR to add something to the CLAUDE.md as part of the PR. Use the Claude Code Github action (/install-github-action) for this (@danshipper's Compounding Engineering)
3) Most sessions start in Plan mode (shift+tab twice). If my goal is to write a Pull Request, I will use Plan mode, and go back and forth with Claude until I like its plan. From there, I switch into auto-accept edits mode and Claude can usually 1-shot it. A good plan is really important.
4) Use Claudes tabs in parallel. Use system notifications to know when a Claude needs input.
5) You can point Claude to look at open source code or any code base you are interested in and learn from it. Eg better patterns for implementing X kind of system
6) Use coding agent to find holes in your design, and in its own design. Iterate at least twice.
7) Clear written requirements help agents build better code.
8) Generated code needs to be readable by engineers, steer agent away from unreadable/uncommon language usages that while optmial ofuscate the code.
	"There is no playbook for building AI native companies, and if there is, I dont have it" ~ Dan Shipper, https://www.youtube.com/watch?v=MGzymaYBiss

	But, here are some good practices learned so far:

	1) One approacj is to use a single Claude.md in the repo.
	Note patterns you want AND patterns you do not want Claude to use as you develop in it.
	Should Contain setup necessary to run the repo, to be used by new engineers (Reduces barried of entry)
	Every member of the team contributes to it from lessons learned.
	2) Tagging @.claude on a PR to add something to the CLAUDE.md as part of the PR. Use the Claude Code Github action (/install-github-action) for this (@danshipper's Compounding Engineering)
	3) Most sessions start in Plan mode (shift+tab twice). If my goal is to write a Pull Request, I will use Plan mode, and go back and forth with Claude until I like its plan. From there, I switch into auto-accept edits mode and Claude can usually 1-shot it. A good plan is really important.
	4) Use Claudes tabs in parallel. Use system notifications to know when a Claude needs input.
	5) You can point Claude to look at open source code or any code base you are interested in and learn from it. Eg better patterns for implementing X kind of system
	6) Use coding agent to find holes in your design, and in its own design. Iterate at least twice.
	7) Clear written requirements help agents build better code.
	8) Generated code needs to be readable by engineers, steer agent away from unreadable/uncommon language usages that while optmial ofuscate the code.
No results found