Your result
YOUR ASSESSMENT
Your result is being forged.
Counted from your own task episodes, chances and all.
What we know
Proof Over Plausible Claims
When an agent reports a finished task or misdiagnoses a failure, you bring concrete proof instead of vague re-requests. You paste exact logs, screenshots of unchanged screens, or status output at the exact moment a decision hinges on them.
This keeps the work anchored to the original goal without getting sidetracked by superficial patches.
Clear Intent and Guarded Action
You consistently name the intended end state and business context up front, giving the agent room to choose the best route. At the same time, you reserve irreversible moves like production merges and data repairs for yourself.
When a repeat failure occurs, you turn manual checks into automated gates, ensuring the same defect cannot ship twice.
Defining Success Before Building
Your next growth move is to state falsifiable pass criteria before implementation starts. Running an observable check after final edits ensures every completion claim passes verification in a live environment before landing.
MEGA Student? This card is not on HQ. Open hq.mega.dev, copy the assessment prompt, run it in your agent, and press Add it to my deck on the link it gives you.