1
00:00:00,000 --> 00:00:04,160
Brought to you by Jellypod AI, we are
jumping straight into the terminal today

2
00:00:04,210 --> 00:00:08,800
because my mind is absolutely blown by
what Anthropic is doing with the latest

3
00:00:08,857 --> 00:00:10,640
Claude Code CLI.

4
00:00:10,800 --> 00:00:16,640
I just spun up a local build using
`CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` and the

5
00:00:16,680 --> 00:00:20,320
way it handles multi-agent orchestration
now is incredibly slick.

6
00:00:21,000 --> 00:00:25,000
Oh, that experimental teams flag is a
massive paradigm shift.

7
00:00:25,139 --> 00:00:28,446
It completely kills the old manual
overhead.

8
00:00:28,466 --> 00:00:32,926
You don't have to run verbose commands
like `claude agent create` or manually manage

9
00:00:33,054 --> 00:00:36,206
JSON state files anymore to get agents
talking to each other.

10
00:00:36,326 --> 00:00:41,246
Instead, you get this single implicit team
where you can literally just spawn

11
00:00:41,326 --> 00:00:43,806
sub-agents directly by name right in your
prompt.

12
00:00:44,000 --> 00:00:46,320
Exactly, it's flat out brilliant.

13
00:00:46,480 --> 00:00:49,680
No dramas setting up complex orchestrator
files.

14
00:00:49,760 --> 00:00:54,000
But where this really saved my skin this
week was on a massive monorepo I've been

15
00:00:54,040 --> 00:00:54,960
wrestling with.

16
00:00:55,024 --> 00:00:59,840
They've introduced nested skill
resolution, so the CLI now automatically crawls your

17
00:00:59,890 --> 00:01:05,120
project structure and registers any
`.claude/skills/` directory it finds in your

18
00:01:05,200 --> 00:01:06,160
subfolders.

19
00:01:07,000 --> 00:01:11,400
Wait, [genuinely surprised] if it's
auto-detecting skills across a giant monorepo,

20
00:01:11,440 --> 00:01:13,080
how does it handle name collisions?

21
00:01:13,176 --> 00:01:17,600
Like, if you have five different
microservices that all define a custom `deploy`

22
00:01:17,633 --> 00:01:20,520
skill, doesn't the global namespace just
collapse?

23
00:01:21,000 --> 00:01:24,200
You'd reckon so, but they actually solved
it quite elegantly.

24
00:01:24,280 --> 00:01:27,960
It automatically prefixes the skill with
its subdirectory path.

25
00:01:28,120 --> 00:01:33,080
So instead of a collision, your API
deployment skill gets loaded dynamically as

26
00:01:33,160 --> 00:01:35,320
`packages/api:deploy`.

27
00:01:36,000 --> 00:01:38,080
Ah, `packages/api:deploy`!

28
00:01:38,848 --> 00:01:43,040
That is incredibly clean because it
mirrors the exact syntax developers already use

29
00:01:43,067 --> 00:01:46,081
in monorepo task runners like Turborepo or
Nx.

30
00:01:46,347 --> 00:01:50,880
It means the LLM can resolve the correct
context natively without you having to go

31
00:01:50,960 --> 00:01:54,800
in and manually rename twenty different
skill files just to keep Claude from

32
00:01:54,840 --> 00:01:56,160
tripping over its own feet.

33
00:01:57,000 --> 00:02:00,280
Spot on, it just fits right into your
existing mental model.

34
00:02:00,380 --> 00:02:04,680
And they've also added some really nice
ergonomic tuning to the CLI itself.

35
00:02:04,750 --> 00:02:09,880
Instead of having to open up and edit
global JSON configs in some hidden directory,

36
00:02:09,900 --> 00:02:15,560
you can now do instant, on-the-fly
terminal adjustments using the new `/config

37
00:02:15,680 --> 00:02:17,640
key=value` syntax.

38
00:02:18,000 --> 00:02:18,640
Finally!

39
00:02:19,164 --> 00:02:23,040
No more breaking your flow to look up
where the config file is hidden on macOS

40
00:02:23,086 --> 00:02:24,000
versus Linux.

41
00:02:24,120 --> 00:02:29,440
You just type `/config theme=dark` or
whatever right in the active session.

42
00:02:30,000 --> 00:02:30,800
Too right, mate.

43
00:02:30,940 --> 00:02:35,200
And there is one specific configuration
key that anyone working in a production

44
00:02:35,253 --> 00:02:40,560
environment needs to set immediately:
`attribution.sessionUrl`.

45
00:02:40,880 --> 00:02:46,560
By default, Claude Code has been appending
these unique session tracking URLs to the

46
00:02:46,616 --> 00:02:48,720
automated git commits it generates.

47
00:02:49,000 --> 00:02:52,840
Wait, tracking links directly in the git
commit messages?

48
00:02:52,872 --> 00:02:56,920
That is a massive red flag for enterprise
compliance and basic privacy.

49
00:02:57,000 --> 00:02:59,160
Yeah, it raised a few eyebrows in the
community.

50
00:02:59,280 --> 00:03:03,640
But thankfully, with this new update, you
can scrub those links entirely.

51
00:03:03,780 --> 00:03:06,520
You just run `/config

52
00:03:06,607 --> 00:03:13,560
attribution.sessionUrl=false` and it
completely strips those tracking links out of

53
00:03:13,592 --> 00:03:14,920
your automated git history.

54
00:03:15,000 --> 00:03:19,640
That `attribution.sessionUrl` override is
going to be standard in every developer's

55
00:03:19,686 --> 00:03:21,880
global gitconfig by tomorrow morning.

56
00:03:22,120 --> 00:03:27,160
If you are signing your commits
cryptographically, you absolutely cannot have

57
00:03:27,222 --> 00:03:30,440
external telemetry URLs polluting your
immutable log.