Claude Code 2.1.286: The verify Skill Before Commit

0:00 This episode's brought to you by Jellypod AI, and we're diving straight in. Ever had an AI agent tell you "all tests pass" and then you find out it never actually ran them? Constantly. And that's the itch Claude Code 2.1.286 scratches. It dropped on September 30, 2026. Go on then. So, according to the changelog, when your project or user skills include one named verify, Claude is now told to run it right before committing.

0:29 Except for docs only and tests only commits. Hang on. That's it? A skill with a particular name? That's the whole feature? That's the whole feature. And I think that's the clever part. Huh. Okay, but why's that better than me just writing "always test before you commit" in CLAUDE.md? I've done that. Mate, I've done that, um, a lot.

0:51 Sure, and it works until it doesn't. Long session, lots of momentum, the context gets compacted, and that one line is, uh, it's just one sentence among thousands. I can't promise that's exactly what happens every time, but that's the failure mode people worry about. A named skill is a fixed hook the tool itself knows about, so it doesn't depend on the model remembering your note.

1:14 Right. So what about a plain old git pre commit hook? That's been around forever. Hooks are great, but they're blind. The commit fails, and the agent sees an error after the fact. With verify, the way I read it, the check runs inside the loop. Claude sees the failure output in context, fixes the typo or the syntax error, runs it again, and only commits once it's clean.

1:37 Ah, so it's less a bouncer at the door and more a mate checking your work before you leave the house. Yeah, that's better than my version. Okay, so what would setup look like? Walk me through it. You'd create a skill folder at .claude/skills/verify in your project. Or put it in ~/.claude/skills/verify if you want it across every project.

2:00 Inside, you describe the pipeline. For a JavaScript project, that might be npm run typecheck, then lint, then npm test run in band so the tests don't fight each other. For Rust, cargo test quietly, then cargo clippy. And then you just, uh, test it. Something like "fix the auth route and commit." Exactly that kind of prompt.

2:22 You should see Claude run verify before the commit. And if you ask it to touch only markdown, it should skip the run, because of that docs only exemption. Which makes sense. Nobody needs a type check for a typo in a README. Right. Now, I've got a worry. I once pushed a bad midnight update that nearly tanked a client's site, so I'm, I'm, I'm a bit twitchy about slow checks.

2:46 What if verify takes forever? Then it's a problem. My take, and this is my opinion, not the changelog's: keep verify fast. If you've got an integration suite that takes ten minutes, don't put it in there. A turn can time out, and you'll be sitting there watching a spinner. Typecheck, lint, the quick tests.

3:06 Leave the heavy stuff for CI. Fair. Fast and boring wins. Always. Oh, and the point release has a bunch of polish too. Permission prompts now show a count like "2 of 5" when several requests stack up. Oh, thank goodness. I hate not knowing how deep the pile is. There's also a safety one. If you're viewing a background agent's or teammate's transcript and type compact, clear, or rewind, it used to silently act on the main conversation.

3:32 Now a dialog names the target and asks first. So you can't accidentally wipe the wrong session. That's, wow, that's a nasty one to have fixed. And in a subagent's view, control enter now moves its running command to the background, so your message gets read right away. Plus, if the API refuses the model your default resolves to, Claude Code retries once on the previous model of the same tier, instead of failing every single turn.

4:00 So the big idea is just, name a skill verify, and the agent checks its own homework before it hands it in. That's it. One folder, one name, and the commit gets a gatekeeper that's actually in the room.