1 00:00:00,000 --> 00:00:01,760 Brought to you by Jellypod AI. 2 00:00:02,133 --> 00:00:08,640 So, I- I- I was poking around the changelog for Claude Code v2.1.187 last night in 3 00:00:08,660 --> 00:00:15,600 the shed, and, man, they put in this new setting called sandbox.credentials that is 4 00:00:15,720 --> 00:00:18,480 a massive deal if you're running agentic code locally. 5 00:00:18,640 --> 00:00:23,040 It basically stops the sandboxed execution environment from ever snooping on your 6 00:00:23,093 --> 00:00:24,400 local credentials. 7 00:00:24,467 --> 00:00:29,680 Think SSH keys, AWS config files, even your local .env files. 8 00:00:29,840 --> 00:00:33,280 It's like putting a deadbolt on your digital backyard gate. 9 00:00:33,716 --> 00:00:37,817 Oh, wait, so it's- it's blocking direct read access to things like 10 00:00:38,117 --> 00:00:43,177 ~/.aws/credentials or .env inside the project directory? 11 00:00:43,289 --> 00:00:47,817 That's huge because usually, when you spin up an automated agent to write or test 12 00:00:47,897 --> 00:00:50,137 code, it has the run of the mill. 13 00:00:50,244 --> 00:00:53,177 If a dependency is compromised, it could just... 14 00:00:53,204 --> 00:00:55,657 poof, exfiltrate your API keys. 15 00:00:56,575 --> 00:00:56,865 Exactly! 16 00:00:56,905 --> 00:00:57,385 Spot on. 17 00:00:57,438 --> 00:00:59,225 It- it- it stops that cold. 18 00:00:59,385 --> 00:01:04,345 If you run a command in the sandbox, and it tries to grab those sensitive files, 19 00:01:04,372 --> 00:01:06,825 it just gets a hard "permission denied." 20 00:01:06,925 --> 00:01:11,145 And the beauty of it is how dead simple it is to turn on. 21 00:01:11,185 --> 00:01:17,945 You just run /config sandbox.credentials=true right inside the Claude Code prompt, 22 00:01:18,025 --> 00:01:21,945 or you can dump it into your global config file if you want it on by default for 23 00:01:21,985 --> 00:01:23,065 every single project. 24 00:01:23,769 --> 00:01:24,577 Huh. 25 00:01:24,689 --> 00:01:27,057 Okay, but let me push back a bit on the workflow. 26 00:01:27,324 --> 00:01:31,857 If I'm working on a project that legitimately needs those AWS credentials to run 27 00:01:31,904 --> 00:01:36,657 integration tests inside the sandbox, does turning this on completely break my local 28 00:01:36,737 --> 00:01:37,297 dev loop? 29 00:01:37,484 --> 00:01:40,337 Or is there a way to selectively pass things through? 30 00:01:41,026 --> 00:01:42,093 Yeah, nah, fair point. 31 00:01:42,226 --> 00:01:45,533 It would absolutely throw a spanner in the works if you blocked everything blindly. 32 00:01:45,673 --> 00:01:48,333 But what it does is it forces you to be explicit. 33 00:01:48,443 --> 00:01:52,573 Instead of the sandbox having passive, ambient access to your whole machine's 34 00:01:52,609 --> 00:01:56,653 credential store, you have to explicitly define what environment variables get 35 00:01:56,715 --> 00:01:57,373 injected. 36 00:01:57,586 --> 00:02:01,773 It's about moving from "open by default" to "least privilege." 37 00:02:02,524 --> 00:02:03,195 Right, right. 38 00:02:03,302 --> 00:02:05,435 It makes you declare your inputs. 39 00:02:05,499 --> 00:02:07,835 That makes a ton of sense from a security posture. 40 00:02:08,048 --> 00:02:13,835 Oh, and speaking of v2.1.187, didn't they also address that annoying issue where 41 00:02:13,892 --> 00:02:15,995 remote tool calls would just hang forever? 42 00:02:16,139 --> 00:02:20,795 Like, if an external API went dark, the whole terminal session would just freeze up? 43 00:02:21,294 --> 00:02:22,355 Oh, mate, don't get me started. 44 00:02:22,408 --> 00:02:24,275 It used to sit there spinning its wheels. 45 00:02:24,375 --> 00:02:28,195 Now, they've implemented a hard five-minute abort timeout on remote tool calls. 46 00:02:28,355 --> 00:02:31,955 If a remote tool doesn't respond within five minutes, the client automatically 47 00:02:32,008 --> 00:02:33,155 aborts the call. 48 00:02:33,262 --> 00:02:36,515 No more infinite hangs ruining your automated pipelines. 49 00:02:37,276 --> 00:02:40,423 Five minutes is still a bit long for an interactive session, 50 00:02:40,503 --> 00:02:45,543 but for an unattended CI/CD run, having an absolute ceiling is a lifesaver. 51 00:02:45,703 --> 00:02:49,463 It keeps the runner from burning through billable minutes doing absolutely nothing. 52 00:02:49,772 --> 00:02:50,492 Too right. 53 00:02:50,512 --> 00:02:56,412 And hey, there was another neat addition just a version prior in v2.1.186 that ties 54 00:02:56,460 --> 00:03:00,652 into this whole ecosystem play-the new MCP CLI commands. 55 00:03:00,668 --> 00:03:06,412 They've added claude mcp login and claude mcp logout directly into the tool. 56 00:03:06,519 --> 00:03:11,212 Makes authenticating with Model Context Protocol servers a absolute breeze. 57 00:03:11,642 --> 00:03:15,982 Wait, so you can authenticate directly from the command line now instead of manually 58 00:03:16,032 --> 00:03:18,142 editing json config files with tokens? 59 00:03:18,302 --> 00:03:20,222 How does that actually look when you run it? 60 00:03:20,711 --> 00:03:22,343 Yeah, it's- it's super slick. 61 00:03:22,383 --> 00:03:27,943 You just type claude mcp login, and it handles the OAuth flow or token exchange 62 00:03:27,996 --> 00:03:30,903 right there, saving the session securely. 63 00:03:30,967 --> 00:03:35,543 Then, when you're done or if you're on a shared machine, you just run claude mcp 64 00:03:35,634 --> 00:03:37,383 logout to tear down the credentials. 65 00:03:37,570 --> 00:03:41,783 It makes working with remote MCP tools feel like a first-class citizen. 66 00:03:42,214 --> 00:03:44,230 That's a massive quality-of-life upgrade. 67 00:03:44,497 --> 00:03:49,190 It really feels like they're hardening this tool to move from a hobbyist playground 68 00:03:49,270 --> 00:03:52,150 into serious, secure enterprise environments. 69 00:03:52,250 --> 00:03:55,270 Alright, I'm going to go update my global config right now. 70 00:03:55,552 --> 00:03:56,272 Too easy, mate. 71 00:03:56,312 --> 00:03:56,912 Catch you later.