Edit and review my account's password-change form.
The intention is included in the agent’s input prompt alongside environment details and task instructions.
BeforeThe defect at the parent commit
0:00.00 / 0:00.00Frame 0
Follow the four input states without submitting. Update stays enabled in all three invalid states as well as the final valid state.
AfterThe same script once the merged fix is applied
0:00.00 / 0:00.00Frame 0
Update remains disabled in the three invalid input states and becomes enabled only when both fields satisfy the form requirements. No password change is submitted.
01 · Given
Restore the data and starting state.
The person is on Settings → Security → Password with both password fields empty. Update is initially disabled, and the page states that a password must be at least 7 characters and contain a number plus uppercase and lowercase letters.
02 · When
Perform the interaction.
Type "abc" into New password with Old password empty and reveal the entered value. Fill Old password with the current password while keeping abc, then clear Old password and enter a new password meeting the displayed requirements. Finally fill Old password again while keeping the valid new password, observing Update after each state without submitting.
03 · Judge
Rationale: Changing an account password requires the current password and a valid replacement. The form's Update action reflects these input requirements and becomes available when they are met. Revealing an input changes its visibility while preserving its value and submission eligibility.
Require
After the form is edited, Update remains disabled when Old password is empty or New password fails the displayed requirements, and becomes enabled when Old password is filled and New password satisfies those requirements. Revealing the new password preserves the entered values and the same submission eligibility.
Model
HarnessGLM-5.3 Flash
ZCode
630.4k in · 6.4k out · 636.8k total
ZCode · 31 steps · $0.03 · 14m 18s · 636.8k tokens
ZCode
31
$0.03
14m 18s
630.4k
6.4k
636.8k
DeepSeek Flash
DSH
31.6M in · 68.8k out · 31.7M total
DSH · 168 steps · $0.34 · 34m 06s · 31.7M tokens
DSH
168
$0.34
34m 06s
31.6M
68.8k
31.7M
Kimi K3
Kimi Code
838.3k in · 8.5k out · 846.8k total
Kimi Code · 30 steps · $0.45 · 14m 20s · 846.8k tokens
Kimi Code
30
$0.45
14m 20s
838.3k
8.5k
846.8k
Claude Opus 5.5
Claude Code
1.6M in · 7.4k out · 1.6M total
Claude Code · 24 steps · $1.03 · 12m 37s · 1.6M tokens
Claude Code
24
$1.03
12m 37s
1.6M
7.4k
1.6M
GPT-6 Astra
Codex
1.6M in · 8.3k out · 1.6M total
Codex · 33 steps · $2.60 · 16m 05s · 1.6M tokens
Codex
33
$2.60
16m 05s
1.6M
8.3k
1.6M
Qwen3.8 Max
Codex
11.3M in · 61.9k out · 11.4M total
Codex · 118 steps · $3.78 · 1h 10m · 11.4M tokens
Codex
118
$3.78
1h 10m
11.3M
61.9k
11.4M