In partnership with
AI stopped halfway again?
WTF is /goal? : A new coding-agent trick. It lets AI keep working for a long time and stop only when the goal is actually done.
Everyone knows this pain.
You ask AI to do a task.
It gives a decent first answer.
Then it stops right before the useful part.
So you type:
continue
Then again.
Then you become the project manager for a machine that was supposed to save your time.
That is the problem Goals are trying to fix.
I am Alex, welcome to ShortCu8 by Innov8.
Lets Dive Deep ๐ฐ
โญToday's Shortcut
Stop giving AI vague prompts.
Give it a finish line.
Normal prompt:
Make this better.
Goal:
Keep working until this exact thing is true.
Use this formula
/goal Do [task].
Done when [measurable result].
Prove it by [specific check].
Stop if [blocked or over budget].
That one line changes the job.
The AI is no longer just replying.
It is working toward a condition.
Why Tools Are Adding Goals
AI agents are not just chatbots anymore.
They are becoming workers.
And workers need:
a target
a test
a stopping point
That is why tools like Codex, Claude Code, and Hermes Agent are adding /goal systems.
Different tools.
Same idea:
Make AI keep going until the outcome is done.
Bad Prompt vs Good Goal
Most people write a bigger prompt.
But the agent needs a clearer scoreboard.
Video Script
Bad:
Make my video script better.
Good:
Rewrite this script until the hook is under 12 words, the intro is clear, and the CTA has one action.
Website
Bad:
Fix my website.
Good:
Check every page, fix broken links, improve mobile layout, and stop when all pages load without console errors.
Research
Bad:
Research AI agents.
Good:
Find 5 beginner-friendly AI agents. Compare price, setup difficulty, use case, and one limitation.
The good version is not fancy.
It just makes "done" visible.
The Chris Hayduk Playbook
Chris Hayduk explained Codex Goals really well.
His main point:
Goal mode is a loop.
The agent acts, Checks progress, Then continues or stops.
If the goal is vague, the loop breaks.
It either stops too early or keeps making random changes.
His 3 rules:
1. Make the goal measurable.
"Better" is weak.
"20 percent faster with tests passing" is usable.
2. Keep feedback fast.
Give the agent a quick test, checklist, preview command, or small dataset.
3. Give it tracking files.
Use:
PLAN.md
ATTEMPTS.md
NOTES.md
This keeps long tasks from turning into memory soup.
Write Your First Goal
Use this:
/goal Complete [task].
Read [files/links] first.
Done when [specific result] is true.
Prove it by [test/checklist/table/screenshot].
Do not change [constraints].
Stop if [blocked].
For Coding
/goal Fix the auth bug.
Done when npm test -- auth exits 0.
Do not change unrelated files.
For Research
/goal Find 7 beginner AI automation tools.
Compare price, setup difficulty, use case, and limitation.
Stop when the table is complete.
The trick is not sounding smart.
The trick is making the stop condition clear.
When Not To Use Goals
Goals are useful, But not for everything.
Do not use them for tiny one-shot tasks, Do not use them when you cannot define success.
Do not use them for risky account actions unless approvals are clear.
And please do not write:
Make everything better.
That is not a goal.
That is a prayer.๐
Honest Verdict
Prompts are still useful.
But prompts alone are weak for long work.
If you want AI to do the grind, give it a goal.
Task. Proof. Stop rule.
That is how you make AI do เดชเดเตเดเดฟ เดชเดฃเดฟ without babysitting it every five minutes.
/goal + Auto mode
Now go build something.
The ShortList
๐ ๏ธCool Tools of the Week:
Claude Platform on AWS: Customers of AWS now have access to all of Claudeโs API features, with AWS authentication, billing, and commitment retirement
ChatGPT add-ons: OpenAIโs model is now available as an add-on for Excel and Google Sheets
Claude Code: Anthropicโs flagship coding platform now has โagent view,โ allowing users to dispatch Claude agents into multiple sessions at once
Snapseed 4.0: Google has rolled out a major update to its photo editing tool for Android.
๐ฉ Innathe Shortcu8 engane undarunnu ๐๏ธ?
We read every reply - just reply to this email and let us know how we can improve !
Appo adutha Shortcu8il kanaam bieโฆ๐
If you read till here, you might find this interesting
#AD1
Fast browsing. Faster thinking.
Your browser gets you to a page. Norton Neo gets you to the answer. The first safe AI-native browser built by Norton moves with you from idea to action without slowing you down. Magic Box understands your intent before you finish typing. AI that works inside your flow, not beside it. No prompting. No copy-pasting. No switching apps.
Built-in AI, instantly and for free. Privacy handled by Norton. Built-in VPN and ad blocking protect you by default. No configuration. No extra apps. Nothing to think about.
Fast. Safe. Intelligent. That's Neo.
#AD2
Works inside Cursor, Warp, VS Code, and every IDE.
Wispr Flow sits at the system level โ dictate into any editor, terminal, or app with full syntax accuracy. No plugins needed. No setup per tool. 89% of messages sent with zero edits.