Review and submit results
Review and submit individual findings or agent-staged outputs, handle failures, and inspect submission history.
Use single-result review when you are inspecting one finding or output record. Use staged review when an agent has collected matching outputs for a reviewer. Agents can stage outputs, but they cannot send them to an external destination.
Submit one result
Section titled “Submit one result”- Open Agents → Output and select the project that produced the result.
- Select an output type, then select the result.
- Choose Submit to….
- Select an enabled action that matches the output type.
- Review the rendered payload and warnings.
- Edit fields marked Editable, then choose Submit.
The Submit to… menu lists enabled actions that match the output’s item_type and point to
an active connection. Each menu option shows the action name, destination type, connection
name, and the last submission status for that output and action.
If no action matches, open Org Settings → Integrations, select the project in Projects, and add an enabled action for the exact output type from Actions.
Preview and edit the payload
Section titled “Preview and edit the payload”Selecting an action opens the preview modal. The modal shows:
- action, destination type, and connection name
- rendered payload fields
- which fields are editable for that adapter
- warnings for unresolved template paths or unverified findings
- the number of prior successful submissions for the same output and action
Only fields marked Editable can be changed. If the original rendered value is an object or array, the edited value must be valid JSON.
Confirm duplicate submissions
Section titled “Confirm duplicate submissions”A submission warns when the output has already succeeded through the same action. To create another report, issue, submission, or webhook call, confirm the duplicate before Dreadnode sends the payload.
Failed attempts do not require duplicate confirmation. If you only need to update an existing external issue, follow its external reference instead of submitting again.
Stage output for human review
Section titled “Stage output for human review”An action with Allow agents to stage lets an agent add matching outputs to the project’s Staged Findings queue. Despite the page name, the queue supports findings, assets, and capability-defined output types.
Staging creates no outbound request. It does not decrypt connection credentials, change the output status, or mark the output as submitted. Repeated staging for the same active output and action returns the existing stage instead of creating duplicate queue entries.
Reviewers can dismiss a stage without submitting it. Dismissal has no confirmation or restore action in the app. Submitted and dismissed stages remain available through the stage-status filter.
Review staged findings in bulk
Section titled “Review staged findings in bulk”Open Staged Findings and select the project. The queue defaults to the finding item type;
choose All types to include assets and capability-defined output types. You can also filter by action,
severity, finding status, stage status, and staging date.
A bulk selection can contain at most 50 stages and must use one submission action. The app stops accepting additional rows when the selection reaches the limit.
- Select staged outputs for one action.
- Choose Review N selected to render every payload.
- Inspect warnings and each rendered payload.
- Continue to confirmation and review every final payload.
- Choose Submit N findings to create one submission attempt per selected stage.
Edits in the bulk JSON editor apply to adapter-approved fields. The confirmation step shows the exact payload snapshot Dreadnode sends and records for each selected stage.
The result panel shows each success and failure. Successful stages move to Submitted. Failed stages stay Staged, so Retry failed can preview and submit only those rows again. Each accepted bulk request receives a batch ID, including retries.
If any selected output has already succeeded through the same action, the confirmation step requires an explicit duplicate acknowledgement. A destination failure for one submission does not stop the rest of the batch.
Account for limits and timeouts
Section titled “Account for limits and timeouts”Each connection accepts up to 60 recorded attempts in a rolling 60-second window. Successful and failed attempts both count. Bulk requests process no more than five outbound calls at once, and each destination call has a 10-second timeout. HackerOne bulk writes run serially below its limit of 25 writes per 20 seconds.
Follow successes and retry failures
Section titled “Follow successes and retry failures”After a successful submission, the modal links to the external record when the destination returns a URL. If the destination only returns an external ID, the modal shows that ID.
The output detail panel keeps successful external URLs under External References. HackerOne and Linear submissions return external URLs. Webhook submissions do not return external references today.
Failed attempts retain the rendered payload, safe error message, destination type, and submitting user. Correct the source output, action template, or saved connection before retrying.
Audit submissions across projects
Section titled “Audit submissions across projects”Organization owners can open Org Settings → Submission Audit to review submissions across every project. Filter by date, project, action, destination, status, or submitter, then open a row to load its submission method, rendered payload, stage provenance, batch ID, safe error, submitting user, and external reference.
The list stays metadata-only and loads detail on demand. The audit never exposes integration credentials, full webhook targets, header values, or provider response bodies.
Trace attempts from the output
Section titled “Trace attempts from the output”The output’s Changelog tab records submission events:
| Event | When it appears |
|---|---|
| Submitted | The destination accepted the payload. |
| Submission failed | The destination rejected the payload, timed out, or could not be reached. |
The changelog shows a short summary for each attempt. Use External References for successful HackerOne or Linear links, and use Submission Audit for the full payload and metadata.