Six tips shared after dogfooding Opus 4.7: auto mode (no more permission prompts), the /fewer-permission-prompts skill, recaps for long-running sessions, focus mode, adaptive-thinking effort levels, and a /go skill pattern for self-verification. Co-Authored-By: Claude <noreply@anthropic.com>
5.2 KiB
6 Tips for Getting More Out of Opus 4.7 — From Boris Cherny
A thread of tips shared by Boris Cherny (@bcherny), creator of Claude Code, on April 16, 2026 — after dogfooding Opus 4.7 for the last few weeks.
| ← Back to Claude Code Best Practice |
Context
After dogfooding Opus 4.7 for a few weeks, Boris has been feeling "incredibly productive" and shared six ways to get more out of the new model — from permission automation to effort tuning to verification patterns.
1/ Auto Mode — No More Permission Prompts
Opus 4.7 loves doing complex, long-running tasks: deep research, refactoring code, building complex features, iterating until a performance benchmark is hit. In the past, you either had to babyset the model while it did these sorts of long tasks, or use --dangerously-skip-permissions.
Anthropic recently rolled out auto mode as a safer alternative. In this mode, permission prompts are routed to a model-based classifier that decides whether the command is safe to run:
- If it's safe, auto-approve
- If it's risky, pause and ask
This means no more babysitting while the model runs. More than that, it means you can run more Claudes in parallel — if safe, you can switch focus to the next Claude.
Auto mode is now available for Opus 4.7 for Max, Teams, and Enterprise users. Shift+Tab to cycle between Ask permissions → Plan mode → Auto mode in the CLI, or choose it from the dropdown in Desktop or VS Code.
2/ The New /fewer-permission-prompts Skill
Anthropic released a new /fewer-permission-prompts skill. It scans through your session history to find common bash and MCP commands that are safe but repeatedly prompt for permission. It then recommends a list of commands to add to your permissions allowlist.
Use this to tune up your permissions and avoid unnecessary permission prompts, especially if you don't use auto mode.
3/ Recaps
Anthropic shipped recaps earlier this week, to prep for Opus 4.7. Recaps are short summaries of what an agent did and what's next.
Very useful when returning to a long-running session after a few minutes or a few hours:
* Cogitated for 6m 27s
* recap: Fixing the post-submit transcript shift bug. The styling-flash
part is shipped as PR #29869 (auto-merge on, posted to stamps). Next:
I need a screen recording of the remaining horizontal rewrap on `cc -c`
to target that separate cause. (disable recaps in /config)
Disable recaps in /config if you don't want them.
4/ Focus Mode
Boris has been loving the new focus mode in the CLI, which hides all the intermediate work to just focus on the final result. The model has reached a point where he generally trusts it to run the right commands and make the right edits. He just looks at the final result.
Use /focus to toggle on/off.
5/ Configure Your Effort Level
Opus 4.7 uses adaptive thinking instead of thinking budgets. To tune the model to think more or less, tune effort.
- Lower effort — faster responses and lower token usage
- Higher effort — the most intelligence and capability
The slider presents five levels: low · medium · high · xhigh · max — Speed on the left, Intelligence on the right.
6/ Give Claude a Way to Verify Its Work
Finally, make sure Claude has a way to verify its work. This has always been important — now 4.7 is 2-3x what you get out of Claude, so it's more important than ever.
Verification looks different depending on the task:
- Backend work — have Claude run your server/service to test end-to-end
- Frontend work — use the Claude Chromium extension to give Claude a way to control your browser
- Desktop apps — use Computer Use
Boris's prompts these days look like Claude do blah blah /go, where /go is a skill that:
- Tests itself end-to-end using bash, browser, or computer use
- Runs
/simplify - Puts up a PR
For long-running work, verification matters even more — when you come back to a task, you know the code works.






