Help & Security Center

How to Use the Platform &
Zero-Data Retention Policy

Everything you need to connect private Git repositories securely, interpret the Brain Score, and review our formal source code deletion guarantee.

1. Connecting Private Git Repositories

Step-by-step instructions to generate restricted Read-Only access tokens

In the audit form, enter your repository URL (e.g. https://github.com/user/repo) and, in the dedicated field below, your Personal Access Token. There is no need to format complex URLs: authentication is handled entirely in ephemeral RAM.

GitHub Generate Fine-Grained Personal Access Token (Recommended)

  1. Log into GitHub, click your profile icon in top-right → Settings.
  2. In the left sidebar, scroll down to Developer Settings → Personal access tokens → Fine-grained tokens (or visit github.com/settings/tokens).
  3. Click Generate new token, give it a name (e.g. DevWithBrain-Audit) and set an expiration (e.g. 7 days).
  4. Under Repository access, select "Only select repositories" and choose the specific target repo.
  5. Under Permissions → Repository permissions, find Contents and set it strictly to Access: Read-only (never grant write, admin, or secrets permissions).
  6. Click Generate token, copy the generated token (github_pat_...), and paste it into the dedicated Token field.

GitLab Create a Project Access Token

In your GitLab repository, navigate to Settings → Access Tokens. Select role Reporter and permission scope read_repository. Generate and paste the resulting token (glpat-...).

2. Zero-Data Retention Policy & IP Protection

Formal guarantees on proprietary source code handling and IP safety

Post-Audit Shredding

Cloned code resides strictly in an ephemeral scratchpad directory. As soon as the report is finalized, the entire directory is permanently destroyed (shred/rm -rf).

Ephemeral Tokens

Personal access tokens are never persisted in databases, logs, or disk caches. They are passed strictly in-memory during git clone and masked (***) on any error.

Zero-Training on LLMs

Supreme Board deliberations run under enterprise contracts with strict Zero-Data Retention guarantees and explicit prohibitions on training AI models on customer inputs.

What is retained in the platform: Only high-level audit metadata (Run ID, timestamp, project name, Brain Score breakdown, and executive report text) accessible to the authorized user. Zero lines of proprietary code remain on disk.

3. Official "Brain Score" Weights (0 - 100)

Weighted formula calibrated for software projects and vibe coding products

1. Market Viability, IP Clearance & PMF
Pitch-vs-Code alignment, Google Trends momentum, competitor radar, and trademark/domain scans.
35%
2. Code Quality, Technical Debt & Innovation Benchmark
Cyclomatic complexity (Lizard CCN), Hotspot risk matrix (CCN x Git Churn), and technical innovation.
25%
3. Security & Compliance
Credential leak detection (Gitleaks patterns), SAST OWASP rules (Semgrep), and dependency CVEs (Trivy).
20%
4. Architectural Benchmark & Scalability
Layer separation, data model consistency, and clean API contracts.
10%
5. Infrastructure Resilience & FinOps / SRE
Inferred virtual topology (DB pooling, serverless statelessness), runaway cost traps, and live prototype TTFB.
10%
"No Hard Blocks, High-Visibility Warnings" Policy

The platform never blocks an audit prematurely: critical findings (such as exposed credentials, trademark collisions, or FinOps loops) apply a prominent score penalty (-10 points) and display in a high-visibility persistent red banner so founders and investors have immediate transparency on risk.

Return to Dashboard & Launch Audit