@ClaudeDevs

@ClaudeDevs

2026-10-03 — saved image

ClaudeDevs @ClaudeDevs

We're adding a new plugin to Claude Code: You should Know.

It scans Claude's output for important information you might miss to help keep you in the loop.

Enable it with:

/plugin enable cc-plugin-you-should-know@builtin

[attached video frame, 0:30, terminal window titled "claude — acme-shop"]
● Bash(npm test -- checkout)
  └ Tests: 12 passed, 12 total

● Update(src/payments/client.ts)
  └ Added 4 lines, removed 1 line
    38   async function chargeCard(order) {
    39 -   return provider.charge(payload(order))
    39 +   return withRetry(() => provider.charge(payload(order)), {
    40 +     tries: 3, backoffMs: 400, timeoutMs: 15000,
    41 +   })

● Bash(npm test)
  └ Tests: 146 passed, 146 total

● Checkout is more reliable now. Here's what changed:

  - Raised the payment request timeout from 5s to 15s
  - Added automatic retries to the payment call (3 tries, with backoff)
  - Cached the cart total so submitting no longer recomputes tax
  - Moved the inventory check ahead of the payment step
  - Disabled the Pay button while a request is in flight
  - Added 9 tests for slow and failed payment responses

  All 146 tests pass. The changes are in src/checkout and src/payments, and nothing outside those two folders was touched.

> 
? for shortcuts

1:25 PM · Oct 2, 2026 · 1.2M Views

claude codepluginanthropicdeveloper tools

@ClaudeDevs

2026-08-07 — saved image

ClaudeDevs ✔ 🅰 @ClaudeDevs · 3h
We're making this change for two reasons:

1. In our testing, auto mode matched or beat manual permission review on every safety measure we tracked.
2. It makes long-horizon work more viable. Claude runs longer between interruptions, so you can run multi-hour tasks in the background without babysitting permissions.
💬 14   🔁 9   ❤ 456   📊 43K   🔖   ⤴

ClaudeDevs ✔ 🅰 @ClaudeDevs · 3h
One reason we trust it more than manual approval: in a study with 1,053 paid testers, we swapped a permission prompt for a clearly dangerous command (text only, nothing actually ran).

Testers caught it 13.6% of the time, and closer to 5% after 50 prompts. Auto mode blocked the same commands 89% of the time, flat across session length.

[bar chart, titled 'Harmful actions caught — Humans vs. auto mode': Human review 13.6%, Auto mode 89%. Source note: 1,053 paid developers recruited for a controlled study; participants were blind to the specific behavior under test.]

ai safetyclaudeanthropictwitteragentic aipermissions