Skip to content

Security: YemenOpenSource/ComposeLintAI

Security

SECURITY.md

Security Policy & Operational Integrity

The ComposeLintAI team takes the security, reliability, and safety of developer environments and source code seriously.


🛡️ Supported Versions

We release security patches, stability updates, and governance improvements for the following versions:

Version Supported Status
1.1.x ✅ Active (Current stable release branch: v1.1.9)
1.0.x ❌ Deprecated
< 1.0 ❌ Unsupported

🔒 Security Architecture & Guarantees

  1. Local-First Default & Optional Remote Funnel:
    • By default, the Bridge Server (server.py) and Agent WebSocket connect exclusively over local loopback interfaces (127.0.0.1 / ADB port forwarding tcp:8711).
    • Remote WAN tunnel previewing via Tailscale Funnel is strictly opt-in, requiring manual activation via the "Share Remote" toolbar action and token authentication.
    • The server does not expose open public relay endpoints or transmit your source code to external cloud services.
  2. Dynamic Auth Token Protection:
    • Every session generates a cryptographically random authentication token stored locally in .composebridge-state/auth.json.
    • Incoming WebSocket connections must supply a valid authentication header or token to execute file modifications.
  3. Byte-Level Journaling & Safe Rollbacks:
    • Any modification made by the AST Splicer generates a .bridge.bak byte-level backup before applying changes.
    • Pending changes are logged to PatternJournal and can be reverted cleanly at any time (Reject All).
  4. Whitelisted Gradle Execution:
    • The MCP build tools only permit explicitly whitelisted Gradle tasks (e.g. assembleDebug, check, compileKotlin). Arbitrary shell command execution is prohibited.
  5. Zero Internal APIs Guarantee:
    • The IntelliJ plugin adheres strictly to official JetBrains Platform public SDKs, eliminating com.intellij.*.impl.* imports to prevent binary breaks or untracked memory leaks.
  6. Agent Debt Hard-Gate Governance:
    • AI agent prompts are structurally hard-gated with machine-readable file coordinates and reverse-order execution to guarantee idempotent, non-destructive source changes.

⚠️ Operational Security & Code Integrity Notices

1. Mobile Live Editing Operational Boundaries

  • AST Mutation Scope: The Live Editing feature (modifying colors, dimensions, and text directly from a connected mobile device) performs real-time AST transformations directly into source files.
  • Project Structure Sensitivity: Non-standard project architectures (such as proprietary design token wrappers, complex custom builders, or unconventional multi-module layouts) may occasionally trigger anomalous code formatting or syntax placement during live edits.
  • Mandatory Human-in-the-Loop Review: Developers must review all incoming modifications in the Android Studio Changes tab (File Diffs) before accepting them. The atomic Approve / Reject controls exist specifically as an architectural safeguard to maintain 100% code integrity.

2. General Tool Reliability

  • All core analytical features—including IDE Architecture Diagnostics, Agent Debt Governance, Token Duplication Scanners, Token Splitting, and Gradle Remote Verification—operate non-destructively and are fully reliable across all standard Kotlin and Kotlin Multiplatform (KMP) codebases.

3. Reporting Live Editing Anomalies vs. Security Vulnerabilities

  • If you encounter anomalous syntax errors or unexpected behavior while editing colors, dimensions, or text from your phone, please open a public bug report with project context on our GitHub Issues Tracker.
  • Security vulnerabilities (unauthorized access, privilege escalation, token leakage) must not be posted publicly and should follow the private disclosure process below.

🚨 Reporting a Vulnerability

If you discover a security vulnerability in ComposeLintAI, please do not open a public issue.

Instead, please report it via:

  1. GitHub Private Security Advisory: Open an advisory under the Security tab of the repository.
  2. Direct Contact: Send an email to the project maintainers via the GitHub profile contact.

Please include:

  • A description of the issue and potential impact.
  • Clear reproduction steps or proof-of-concept (PoC).
  • Your environment details (OS, ComposeLintAI version, Kotlin version, Python version).

We will acknowledge receipt within 48 hours and provide regular progress updates until a patch is released.

There aren't any published security advisories