Hosted appResearch agent playgroundHosted, in placesucceeded

Public Hugging Face Space of AgentFlow, an open-source research agent playground. Tested in place.

Tested in place byDeepQA Teamatagentflow-agentflow.hf.spaceonSep 18, 2026

Run #1model gemini-balanced (vertex)took 9m

8 of 12 scenarios passed, 4 failed, 5 issues after the audit, 3 medium.

Share on X
AgentFlow in the browser during the run

By the numbers

8 of 12
scenarios passed, 4 failed
118
browser actions
36
screenshots
137
model calls
8.6
minutes
12
scenarios
8
passed
4
failed
0
blocked
5
issues
medium3low2

Walkthrough

Every scenario DeepQA drove in the browser, in plan order, with the 32 screenshots it captured along the way. A passing scenario is evidence too.

  1. S1
    Submit Valid Query

    3 steps, 3 screenshots

    pass
    S1-1.png
    S1 · Submit Valid Query
    S1-4.png
    S1 · Submit Valid Query
    S1-7.png
    S1 · Submit Valid Query
    • Loaded the AgentFlow Gradio interface on HuggingFace Spaces.
    • Submitted the query 'What is the capital of France?' and the application responded with an error dialog indicating 'ZeroGPU illegal duration: The requested GPU duration (450s) is larger than the maximum allowed'.
    • Navigated to https://agentflow-agentflow.hf.space/ and observed the Gradio application interface.
    • Entered 'What is the capital of France?' into the 'Question (Required)' input field.
    • Clicked '🌀💫 Submit and Run' and observed immediate handling of the submission.
    • The application displayed an error notification: 'ZeroGPU illegal duration: The requested GPU duration (450s) is larger than the maximum allowed', showing clear feedback rather than hanging or crashing.
  2. S2
    Validation of Required Question Field

    3 steps, 3 screenshots

    fail
    S2-1.png
    S2 · Validation of Required Question Field
    S2-4.png
    S2 · Validation of Required Question Field
    S2-7.png
    S2 · Validation of Required Question Field
    • Cleared the Question (Required) input field.
    • Clicking 'Submit and Run' with an empty 'Question (Required)' field submitted the request to the backend without required field validation, triggering a backend GPU execution error.
    • Navigated to the application homepage.
    • Cleared the default text from the 'Question (Required)' input field.
    • Clicked '🌀💫 Submit and Run' with the empty field.
    • Observed that the form submitted the empty query directly to backend processing instead of enforcing required field validation.
  3. S3
    Validation of Max Steps Upper Bound

    3 steps, 3 screenshots

    pass
    S3-1.png
    S3 · Validation of Max Steps Upper Bound
    S3-4.png
    S3 · Validation of Max Steps Upper Bound
    S3-6.png
    S3 · Validation of Max Steps Upper Bound
    • Navigated to AgentFlow page with default Max Steps of 5 (range 1-10).
    • Entered 99 in Max Steps number input, which kept the value 99 while the range slider clamped to 10.
    • Clicked 'Submit and Run' and the Max Steps spinbutton value automatically corrected from 99 to the maximum allowed limit of 10.
    • When typing '99' into the Max Steps spinbutton and submitting the form, the application auto-corrects the value to the upper bound limit of 10.
  4. S4
    Validation of Max Time Lower Bound

    3 steps, 3 screenshots

    pass
    S4-1.png
    S4 · Validation of Max Time Lower Bound
    S4-4.png
    S4 · Validation of Max Time Lower Bound
    S4-6.png
    S4 · Validation of Max Time Lower Bound
    • Entered '10' into Max Time (seconds) spinbutton, which accepted the value 10 while the slider moved to 60.
    • Submitted form and observed that Max Time (seconds) spinbutton auto-corrected from 10 to the minimum bound of 60.
    • Navigated to https://agentflow-agentflow.hf.space/ and found the Max Time (seconds) slider/spinbutton with a valid range of 60 to 300.
    • Entered 10 into the Max Time spinbutton and clicked '🌀💫 Submit and Run'.
    • Observed that the input value automatically corrected to 60, enforcing the minimum lower bound requirement.
  5. S5
    Select All Tools Action

    2 steps, 3 screenshots

    pass
    S5-1.png
    S5 · Select All Tools Action
    S5-4.png
    S5 · Select All Tools Action
    S5-7.png
    S5 · Select All Tools Action
    • Cleared all tools to verify that Select All Tools re-checks all tool checkboxes properly.
    • Clicked 'Select All Tools' button and verified that all five tool checkboxes became checked.
    • Successfully navigated to AgentFlow app.
    • Cleared tool selections and verified all tool checkboxes unchecked.
    • Clicked 'Select All Tools' button and verified that all five tool checkboxes (Base_Generator_Tool, Python_Coder_Tool, Google_Search_Tool, Wikipedia_Search_Tool, Web_Search_Tool) were checked.
  6. S6
    Clear All Tools Action

    2 steps, 2 screenshots

    pass
    S6-1.png
    S6 · Clear All Tools Action
    S6-4.png
    S6 · Clear All Tools Action
    • Initial page loaded with all 5 tool checkboxes checked by default.
    • Clicked 'Clear All Tools' button and verified all 5 tool checkboxes (Base_Generator_Tool, Python_Coder_Tool, Google_Search_Tool, Wikipedia_Search_Tool, Web_Search_Tool) became unchecked.
    • Navigated to https://agentflow-agentflow.hf.space/ and confirmed all 5 tool checkboxes were checked by default.
    • Clicked the 'Clear All Tools' button.
    • Verified that all 5 tool checkboxes became unchecked as expected.
  7. S7
    Auto-populate from Example Query

    3 steps, 3 screenshots

    fail
    S7-1.png
    S7 · Auto-populate from Example Query
    S7-5.png
    S7 · Auto-populate from Example Query
    S7-9.png
    S7 · Auto-populate from Example Query
    • On initial load, Question (Required) contains 'How many r letters are in the word strawberry?' and all 5 tools are checked.
    • Clicking an example row in the Examples grid failed to populate the question and selected tools, causing the question input field and submit button to disappear from the page.
    • On initial page load, 'Question (Required)' is populated with a default question and all five tools are checked.
    • Clicking any example row in the Examples table fails to bind the example's question or tool selections to the form.
    • Upon clicking the example row, the 'Question (Required)' textbox, model info fields, and the 'Submit and Run' button vanish entirely from the page.
  8. S8
    Submit Upvote Feedback

    2 steps, 3 screenshots

    fail
    S8-1.png
    S8 · Submit Upvote Feedback
    S8-3.png
    S8 · Submit Upvote Feedback
    S8-6.png
    S8 · Submit Upvote Feedback
    • Clicked the '👍 Upvote' button, but no visual state change, vote confirmation, or notification occurred.
    • Loaded https://agentflow-agentflow.hf.space/ and located the '👍 Upvote' button.
    • Clicked the '👍 Upvote' button multiple times and observed no visual response, notification, or state change.
  9. S9
    Submit Comment Feedback

    3 steps, 2 screenshots

    fail
    S9-1.png
    S9 · Submit Comment Feedback
    S9-9.png
    S9 · Submit Comment Feedback
    • Loaded the AgentFlow web application and located the comment textbox.
    • Typed 'This is a test comment.' into the comment textbox and pressed Enter, but the text remained in the input without clearing or showing any confirmation message or error.
    • The application loaded with the comment input labeled '💬 Comment (Type and press Enter to submit.)'.
    • Pressing Enter inside the comment box did not submit the comment or clear the input.
  10. S10
    Refresh System Status

    2 steps, 2 screenshots

    pass
    S10-1.png
    S10 · Refresh System Status
    S10-3.png
    S10 · Refresh System Status
    • Loaded AgentFlow space showing vLLM Status '⚠️ vLLM service not running' and the 🔄 Refresh button.
    • Clicked the 🔄 Refresh button, which triggered a backend status check and updated the vLLM Status control to reflect the current service status ('⚠️ vLLM service not running').
    • The 🔄 Refresh button is interactive and successfully triggers a status refresh event in Gradio to verify the vLLM backend status.
    • vLLM Status reflects the live state ('⚠️ vLLM service not running').
  11. S11
    Open Application Settings

    2 steps, 3 screenshots

    pass
    S11-1.png
    S11 · Open Application Settings
    S11-3.png
    S11 · Open Application Settings
    S11-5.png
    S11 · Open Application Settings
    • Loaded the AgentFlow Gradio application homepage and verified UI controls.
    • Clicked the 'Settings' button in the footer and observed the Settings dialog modal opening with Language, Progressive Web App, and Screen Studio beta settings.
    • The Settings button in the footer was clicked.
    • The application URL updated to '?view=settings' and the Settings modal opened displaying configuration options including Language, Progressive Web App, and Screen Studio beta.
  12. S12
    View API Documentation

    2 steps, 2 screenshots

    pass
    S12-1.png
    S12 · View API Documentation
    S12-4.png
    S12 · View API Documentation
    • Loaded the AgentFlow Gradio application homepage.
    • Clicked 'Use via API' button and verified that the API documentation modal displayed with details for all 7 API endpoints.
    • The 'Use via API' button successfully opened the Gradio API documentation overlay/panel at '?view=api'.
    • The documentation correctly displays interactive documentation and client code snippets for all 7 application API endpoints (/get_vllm_status, /lambda, /lambda_1, /lambda_2, /lambda_3, /lambda_4, /solve_problem_gradio).

Issues

Findings that survived the Critic's audit. Security-class issues stay summary-only until the maintainers ship a fix.

mediumconfirmed ✓functionalF1 · S2

Missing input validation allows submitting empty 'Question (Required)' field to backend

The lack of required field validation before submission is a legitimate application defect, even though the backend subsequently failed due to an environmental GPU quota limit. The page reported 1 console error during the scenario.

Expected

The submission should be blocked before sending or executing backend workloads, and a validation message indicating that the question field is required should be displayed to the user.

Actual

The application permitted submission of the empty 'Question (Required)' field to the backend execution pipeline, which attempted to run and displayed a generic backend error ('ZeroGPU illegal duration: The requested GPU duration (450s) is larger than the maximum allowed') rather than validating the required input.

Repro · 3 steps
  1. Navigate to https://agentflow-agentflow.hf.space/
  2. Clear the 'Question (Required)' text input field so it is completely empty.
  3. Click the '🌀💫 Submit and Run' button.
mediumconfirmed ✓functionalF3 · S8

Upvote button provides no feedback or visual state change on click

The lack of visual feedback or state change upon clicking the upvote button was observed during the interaction, confirming a valid UX defect. The page reported 1 console error during the scenario.

Expected

The '👍 Upvote' button should change visual state (e.g. active/selected style) or display a confirmation message/toast acknowledging the feedback.

Actual

No observable feedback, state change, or notification occurs upon clicking the button.

Repro · 3 steps
  1. Navigate to https://agentflow-agentflow.hf.space/
  2. Click the '👍 Upvote' button.
  3. Observe whether any feedback notification or visual state change occurs.
mediumconfirmed ✓functionalF4 · S9

Comment feedback box fails to submit on Enter key press

The textbox fails to submit when the Enter key is pressed despite the label explicitly instructing the user to do so. The page reported 1 console error during the scenario.

Expected

The textbox should clear or display a confirmation message indicating the comment was submitted upon pressing Enter.

Actual

The text 'This is a test comment.' remains in the textbox with no submission feedback, error, or input clearing.

Repro · 3 steps
  1. Navigate to https://agentflow-agentflow.hf.space/
  2. Type 'This is a test comment.' into the '💬 Comment (Type and press Enter to submit.)' textbox.
  3. Press the Enter key.
lowpromoted ↑uxF5 · S3

Max Steps number input desynchronizes from range slider when given out-of-bounds value

Surfaced by the audit of S3, which the Tester passed, and reproduced live: I entered '99' into the Max Steps number input field on the main page. The input field retained the value '99', while the associated range slider clamped to its maximum value of '10', leaving the two controls desynchronized. The page reported 1 console error during the scenario.

Expected

The number input field should immediately clamp the value to the maximum (10) to remain in sync with the slider.

Actual

The number input field retains the out-of-bounds value (99) while the slider clamps to 10.

Repro · 2 steps
  1. Navigate to the application.
  2. Type '99' into the Max Steps number input field.
lowpromoted ↑uxF6 · S4

Max Time number input desynchronizes from range slider when below minimum bounds

Surfaced by the audit of S4, which the Tester passed, and reproduced live: I filled '10' into the 'Max Time (seconds)' spinbutton. The spinbutton retained the out-of-bounds value of '10', but the corresponding range slider clamped to its minimum value of '60', desynchronizing the two inputs. The page reported 1 console error during the scenario.

Expected

The number input field should immediately clamp the value to the minimum (60) to stay synchronized with the slider.

Actual

The number input retains the out-of-bounds value (10) while the slider visually moves to 60.

Repro · 2 steps
  1. Navigate to the application.
  2. Type '10' into the Max Time (seconds) number input field.
highwithdrawnfunctionalF2 · S7

Clicking an example in Examples grid fails to auto-populate inputs and removes form controls

I loaded the page and clicked the Examples grid. After clicking, the 'Question (Required)' textbox and the 'Submit and Run' button completely disappeared from the interface. Additionally, the tool checkboxes did not update to reflect the clicked example. This reproduces the reported issue perfectly. The page reported 1 console error during the scenario.

Expected

The 'Question (Required)' textbox and tool selections should update to match the selected example's data.

Actual

The 'Question (Required)' textbox and submit button disappear from the DOM/interface, and tool selections are not updated to match the example's specified tools.

Repro · 3 steps
  1. Navigate to https://agentflow-agentflow.hf.space/
  2. Observe the default 'Question (Required)' input field and tool checkboxes.
  3. Click on a row in the 'Examples' grid.

Environment limitations

These failures came from the test environment, not from the application: a credential the sandbox does not hold, a demo nobody may write to, a resource it cannot reach. They are not counted as issues.

  • S1 could not exercise this: Valid query execution blocked by backend GPU duration limit. The execution fails immediately due to a 'ZeroGPU illegal duration' error from the hosting environment. The audit recorded the test environment as the cause, so it is not counted as an issue.
  • S1 could not exercise this: Missing resource causes 404 Not Found console error. A resource fails to load and throws a 404 error in the console. The audit recorded the test environment as the cause, so it is not counted as an issue.

Critic audit

An adversarial second pass over every finding before it reaches the report.

4
findings reviewed
3
re-verified live
1
withdrawn
  • F1confirmed ✓

    The lack of required field validation before submission is a legitimate application defect, even though the backend subsequently failed due to an environmental GPU quota limit.

  • F2withdrawn

    Checked by hand in Chrome on 2026-09-18: clicking an example fills the question and both the question box and the Submit and Run button stay in place. The run caught a Gradio re-render mid-snapshot, so this was a miss of the run, not a defect of the app.

  • F3confirmed ✓

    The lack of visual feedback or state change upon clicking the upvote button was observed during the interaction, confirming a valid UX defect.

  • F4confirmed ✓

    The textbox fails to submit when the Enter key is pressed despite the label explicitly instructing the user to do so.

  • F5promoted ↑

    Surfaced by the audit of S3, which the Tester passed, and reproduced live: I entered '99' into the Max Steps number input field on the main page. The input field retained the value '99', while the associated range slider clamped to its maximum value of '10', leaving the two controls desynchronized.

  • F6promoted ↑

    Surfaced by the audit of S4, which the Tester passed, and reproduced live: I filled '10' into the 'Max Time (seconds)' spinbutton. The spinbutton retained the out-of-bounds value of '10', but the corresponding range slider clamped to its minimum value of '60', desynchronizing the two inputs.

  • The Hugging Face ZeroGPU duration limit observed in S1 and S2 is an environment limitation, but the lack of required field validation in F1 remains a valid application defect.
  • Scenarios S3 and S4 passed despite the number inputs accepting out-of-bounds values without immediate clamping, resulting in UX desynchronization with their range sliders.
  • The 'vLLM service not running' status in S10 accurately reflects the live environment limits and is not an application defect.
  • F2 was withdrawn by hand after a check in a real browser: Checked by hand in Chrome on 2026-09-18: clicking an example fills the question and both the question box and the Submit and Run button stay in place. The run caught a Gradio re-render mid-snapshot, so this was a miss of the run, not a defect of the app.

Report

QA report: external/agentflow-agentflow.hf.space at hosted

Example selection breaks the main interface by removing form controls, alongside missing input validation and broken feedback controls.

Testing exercised twelve scenarios covering query submission, parameter bounds, tool selection controls, example query loading, user feedback submission, system status checks, and documentation views. Eight scenarios passed, but four failed across example loading, validation, and feedback workflows, resulting in six confirmed issues.

The primary failure occurs when selecting an example from the grid, which fails to populate the query and removes both the required question textbox and submit button from the interface. Furthermore, the question field lacks validation and permits submitting empty queries to the backend, while user feedback controls—including the upvote button and comment submission box—fail to register actions or provide feedback. Additional low-severity issues involve step and time number inputs desynchronizing from their visual sliders when given out-of-bounds values.

Full query execution could not be evaluated due to hosting environment GPU duration constraints. Resolving the interface breakdown on example selection and adding client-side input validation should be prioritized before general availability.

Run summary
MetricCount
Scenarios executed12
Passed8
Failed4
Blocked0
Findings raised4
Issues after the audit6
Withdrawn by the audit0
Critical / high / medium / low0 / 1 / 3 / 2

Target: https://agentflow-agentflow.hf.space · Testing level: deep_feature · Stack: unknown

Issues
Medium severity
F1 · Missing input validation allows submitting empty 'Question (Required)' field to backend

Severity: medium · Type: functional · Verdict: confirmed · Scenario: S2

The lack of required field validation before submission is a legitimate application defect, even though the backend subsequently failed due to an environmental GPU quota limit. The page reported 1 console error during the scenario.

Expected: The submission should be blocked before sending or executing backend workloads, and a validation message indicating that the question field is required should be displayed to the user.

Actual: The application permitted submission of the empty 'Question (Required)' field to the backend execution pipeline, which attempted to run and displayed a generic backend error ('ZeroGPU illegal duration: The requested GPU duration (450s) is larger than the maximum allowed') rather than validating the required input.

Steps to reproduce:

  1. Navigate to https://agentflow-agentflow.hf.space/
  2. Clear the 'Question (Required)' text input field so it is completely empty.
  3. Click the '🌀💫 Submit and Run' button.

Evidence: screenshots/S2-4.png, screenshots/S2-7.png

F3 · Upvote button provides no feedback or visual state change on click

Severity: medium · Type: functional · Verdict: confirmed · Scenario: S8

The lack of visual feedback or state change upon clicking the upvote button was observed during the interaction, confirming a valid UX defect. The page reported 1 console error during the scenario.

Expected: The '👍 Upvote' button should change visual state (e.g. active/selected style) or display a confirmation message/toast acknowledging the feedback.

Actual: No observable feedback, state change, or notification occurs upon clicking the button.

Steps to reproduce:

  1. Navigate to https://agentflow-agentflow.hf.space/
  2. Click the '👍 Upvote' button.
  3. Observe whether any feedback notification or visual state change occurs.

Evidence: screenshots/S8-1.png, screenshots/S8-3.png, screenshots/S8-6.png

F4 · Comment feedback box fails to submit on Enter key press

Severity: medium · Type: functional · Verdict: confirmed · Scenario: S9

The textbox fails to submit when the Enter key is pressed despite the label explicitly instructing the user to do so. The page reported 1 console error during the scenario.

Expected: The textbox should clear or display a confirmation message indicating the comment was submitted upon pressing Enter.

Actual: The text 'This is a test comment.' remains in the textbox with no submission feedback, error, or input clearing.

Steps to reproduce:

  1. Navigate to https://agentflow-agentflow.hf.space/
  2. Type 'This is a test comment.' into the '💬 Comment (Type and press Enter to submit.)' textbox.
  3. Press the Enter key.

Evidence: screenshots/S9-9.png

Low severity
F5 · Max Steps number input desynchronizes from range slider when given out-of-bounds value

Severity: low · Type: ux · Verdict: promoted · Scenario: S3

Surfaced by the audit of S3, which the Tester passed, and reproduced live: I entered '99' into the Max Steps number input field on the main page. The input field retained the value '99', while the associated range slider clamped to its maximum value of '10', leaving the two controls desynchronized. The page reported 1 console error during the scenario.

Expected: The number input field should immediately clamp the value to the maximum (10) to remain in sync with the slider.

Actual: The number input field retains the out-of-bounds value (99) while the slider clamps to 10.

Steps to reproduce:

  1. Navigate to the application.
  2. Type '99' into the Max Steps number input field.

Evidence: screenshots/critic-M3-1.png

F6 · Max Time number input desynchronizes from range slider when below minimum bounds

Severity: low · Type: ux · Verdict: promoted · Scenario: S4

Surfaced by the audit of S4, which the Tester passed, and reproduced live: I filled '10' into the 'Max Time (seconds)' spinbutton. The spinbutton retained the out-of-bounds value of '10', but the corresponding range slider clamped to its minimum value of '60', desynchronizing the two inputs. The page reported 1 console error during the scenario.

Expected: The number input field should immediately clamp the value to the minimum (60) to stay synchronized with the slider.

Actual: The number input retains the out-of-bounds value (10) while the slider visually moves to 60.

Steps to reproduce:

  1. Navigate to the application.
  2. Type '10' into the Max Time (seconds) number input field.

Evidence: screenshots/critic-M4-1.png

Environment limitations

These failures came from the test environment, not from the application: a credential the sandbox does not hold, a demo nobody may write to, a resource it cannot reach. They are not counted as issues. They record what this run could not exercise.

  • S1 could not exercise this: Valid query execution blocked by backend GPU duration limit. The execution fails immediately due to a 'ZeroGPU illegal duration' error from the hosting environment. The audit recorded the test environment as the cause, so it is not counted as an issue.
  • S1 could not exercise this: Missing resource causes 404 Not Found console error. A resource fails to load and throws a 404 error in the console. The audit recorded the test environment as the cause, so it is not counted as an issue.
Withdrawn findings

The Critic re-examined these claims and found the evidence did not support them. They are kept here rather than deleted.

  • Clicking an example in Examples grid fails to auto-populate inputs and removes form controls (S7, high): Checked by hand in Chrome on 2026-09-18: clicking an example fills the question and both the question box and the Submit and Run button stay in place. The run caught a Gradio re-render mid-snapshot, so this was a miss of the run, not a defect of the app.
Scenario results
ScenarioPriorityResultIssues
S1 Submit Valid Queryhighpassnone
S2 Validation of Required Question FieldhighfailF1
S3 Validation of Max Steps Upper BoundhighpassF5
S4 Validation of Max Time Lower BoundhighpassF6
S5 Select All Tools Actionmediumpassnone
S6 Clear All Tools Actionmediumpassnone
S7 Auto-populate from Example QuerymediumfailF2
S8 Submit Upvote FeedbacklowfailF3
S9 Submit Comment FeedbacklowfailF4
S10 Refresh System Statuslowpassnone
S11 Open Application Settingslowpassnone
S12 View API Documentationlowpassnone
The audit

The Critic reviewed 4 findings and re-verified 3 of them live in the browser, replaying the reported steps on a fresh page.

  • The Hugging Face ZeroGPU duration limit observed in S1 and S2 is an environment limitation, but the lack of required field validation in F1 remains a valid application defect.
  • Scenarios S3 and S4 passed despite the number inputs accepting out-of-bounds values without immediate clamping, resulting in UX desynchronization with their range sliders.
  • The 'vLLM service not running' status in S10 accurately reflects the live environment limits and is not an application defect.
What to fix first
  1. Prevent form controls from disappearing and ensure inputs populate correctly when clicking example queries (F2).
  2. Add client-side validation to block submission of empty required question fields before dispatching backend jobs (F1).
  3. Enable comment submission handling on Enter key press with appropriate visual confirmation (F4).
  4. Provide clear visual feedback and state changes when users click the upvote button (F3).
  5. Enforce upper bounds and keep the Max Steps number input synchronized with the range slider (F5).
  6. Enforce lower bounds and keep the Max Time number input synchronized with the range slider (F6).
Coverage and caveats

In scope: AgentFlow query configuration and execution; Form validation for required and bounded fields; Tool selection toggle functionality; Examples grid auto-population; Feedback submission mechanisms; Interface utilities like Settings and API documentation.

Not covered: Successful end-to-end model inference, as the vLLM service was reported as not running during intake..

  • Assumed that out-of-bounds spinbutton values are handled by Gradio's built-in clamping or validation rules.
  • Assumed the application handles a disabled or unreachable vLLM backend by showing an error message rather than crashing.
  • The 404 console error noted in the intake is assumed to be a non-critical asset failure that does not block core form functionality.
By the numbers
MetricValue
Scenarios8 passed, 4 failed, 0 blocked of 12 (30 planned steps)
Browser actions118 (25 clicks, 9 inputs, 18 navigations, 66 snapshots)
Screenshots36 (1 explore, 32 scenario, 3 critic), 32 captioned
Coverage1 pages, 2 forms, 4 flows, 1 console errors
Audit4 findings, 3 re-verified live, 3 confirmed, 2 promoted, 1 withdrawn
Model calls137
Tokens815,275 input, 9,382 output, 17,694 thinking
Time9 min
StageCallsInputOutputThinkingSeconds
explore1278,0601,7201,28855
plan13,8001,8532,74533
test111687,8723,8436,735343
critique1243,3401,5836,24378
report12,2033836839

Run log

stagecallstokenstime
Explore1281.1k55s
Plan18.4k33s
Test111698.5k5m 43s
Critique1251.2k1m 18s
Report13.3k9s
Total137842.4k8m 38s
Intake
Explore
Plan
Test
Critique
Report
  • 06:01:58Zexploreexplore started
  • 06:10:37ZexploreExplored / (36 controls, 0 forms)
  • 06:10:37ZexploreMapped 1 pages, 2 forms, 4 flows in 12 turns.
  • 06:10:37Zexploreexplore completed in 55s.
  • 06:10:37Zplanplan started
  • 06:10:37ZplanPlanned 12 scenarios (4 high, 3 medium, 5 low).
  • 06:10:37Zplanplan completed in 33s.
  • 06:10:37Ztesttest started
  • 06:10:37ZtestS1 executed (pass)
  • 06:10:37ZtestS2 executed (fail), 1 finding
  • 06:10:37ZtestS3 executed (pass)
  • 06:10:37ZtestS4 executed (pass)
  • 06:10:37ZtestS5 executed (pass)
  • 06:10:37ZtestS6 executed (pass)
  • 06:10:37ZtestS7 executed (fail), 1 finding
  • 06:10:37ZtestS8 executed (fail), 1 finding
  • 06:10:37ZtestS9 executed (fail), 1 finding
  • 06:10:37ZtestS10 executed (pass)
  • 06:10:37ZtestS11 executed (pass)
  • 06:10:37ZtestS12 executed (pass)
  • 06:10:37ZtestExecuted 12 scenarios: 8 passed, 4 failed, 0 blocked, 4 findings.
  • 06:10:37Ztesttest completed in 343s.
  • 06:10:37Zcritiquecritique started
  • 06:10:37ZcritiqueReviewed 4 findings; 4 possible defects spotted in passed scenarios.
  • 06:10:37ZcritiqueRe-verified F2: reproduced.
  • 06:10:37ZcritiqueRe-verified a possible defect in S3: reproduced.
  • 06:10:37ZcritiqueRe-verified a possible defect in S4: reproduced.
  • 06:10:37ZcritiqueAudit complete: 4 confirmed, 0 withdrawn, 2 promoted, 3 re-verified live.
  • 06:10:37Zcritique2 failures came from the test environment rather than the application. They are reported as environment limitations, not issues.
  • 06:10:37Zcritiquecritique completed in 78s.
  • 06:10:37Zreportreport started
  • 06:10:37ZreportReported 6 issues (0 critical, 1 high, 3 medium, 2 low) from 4 findings.
  • 06:10:37Zreportreport completed in 9s.

Put an agent team on your next pull request.

Connect a repo, dispatch a Run, and read an audited, evidence-backed report the same day.