Quickstart
The English user guide is currently in beta preview. Most of the documents have been automatically translated from the Japanese version. Should you find any inaccuracies, please reach out to Flatt Security.
This page walks you through one round of Takumi assessment, in the order shown below.
What Takumi Assessments Do
- Even without security expertise, you just provide the source code or the URL, and Takumi performs a comprehensive assessment.
- Takumi lists the features of the application and assesses each one from each perspective, such as injection or broken authorization, starting with the highest-risk combinations.
- A credit threshold caps each run, so you can review the results and add credits step by step.
- From the report, you can go on to fixing the findings, verifying the fixes, and a penetration test, a more advanced form of testing.
Before You Start
You need an organization with an active base subscription, and the Owner, Takumi manager, or Takumi user role in it. What else you need depends on what Takumi assesses.
| What Takumi assesses | What you need |
|---|---|
| Source code | A GitHub integration, or the source code as a .zip or .tar.gz archive |
| Running application | Organization or ownership verification, and accounts for signing in if the application requires it |
Choose an Assessment Method
Choose one of the three methods below. If you can provide both the source code and a running application, choose a graybox assessment.
| Method | What you provide | What Takumi does |
|---|---|---|
| Graybox (recommended) | Source code and a running application | Finds candidates in the code, and reports only those it reproduces in the application |
| Whitebox | Source code | Finds vulnerabilities by analyzing the code |
| Blackbox | A running application | Finds vulnerabilities by attacking the application |
Until your organization creates its first assessment, the creation screen shows a guide to choosing a method. For details, see Choosing an Assessment Method.

Run Your First Assessment
Open Assessment from the sidebar, click Create Assessment, and select Create new assessment. Enter the name and report language, choose the method, select Full Assessment as the assessment type, and set the credit threshold.

Then specify the target as below for the method you chose, and click Start Pentesting.
- Graybox
- Whitebox
- Blackbox
Add the repository, or upload an archive, and enter the URL of the application running that code. If signing in is required, add the credentials under In-App Authentication.

Add the repository, or upload an archive.

Enter the URL to start crawling from. If signing in is required, add the credentials under In-App Authentication.

While the Assessment Runs
An assessment usually takes hours or more. It keeps running after you close the page, so set up Slack, email, or webhook notifications to learn when it stops.

Continue or Complete the Assessment
When the credit threshold is reached, or every combination has been assessed, the assessment stops in Pending.
Opening a pending assessment shows which combinations of feature and perspective have been assessed. On this screen, you can take one of the following three actions.
- Press Preview Report to view a report that summarizes the results of the assessment so far
- Press Start Pentesting to resume the assessment of the remaining combinations within the credit threshold you specify
- Press Complete Assessment to finalize the assessment with the results so far

Review the Report
The report lists the findings in order of urgency. A graybox report lists the findings reproduced in the application apart from those that could not be reproduced. You can download the report as Markdown or PDF.

Act on the Findings
Use the Next actions on the right of the report.
Fix Vulnerabilities
Fix Vulnerability has Takumi open a pull request that fixes each finding you select. This requires a GitHub App with write access to the repository. For details, see Autofix.

Verify the Fix
Once the fix is deployed, Verify Vulnerability checks whether a finding still reproduces, usually within a few minutes. For details, see Vulnerability Verification.

Follow Up with a Penetration Test
From a graybox or blackbox report, Pentest starts a penetration test that carries over the assessment's settings and results. It looks for vulnerabilities the assessment did not detect, and checks whether the detected ones combine into more serious damage. For details, see Carrying Over Results.

Retest the Application
Retest assesses the application again with the same settings, for example after you have fixed several findings.
Next Steps
- Periodic Assessment: assess the files changed in a repository every week or month
- Dependabot PR Auto-Triage: have Takumi triage dependency update pull requests
- Takumi API: run Takumi from CI/CD or your own workflows