Skip to content

Roadmap

Jump to bottom
Varun Pratap Bhardwaj edited this page Mar 3, 2026 · 3 revisions

Roadmap -- SkillFortify Development Plan

This page tracks completed milestones and planned development.

Completed Releases

v0.1.0 -- Foundation (February 2026)

Established the formal foundations for agent skill supply chain security.

  • Formal threat model (DY-Skill attacker model, 5 supply chain phases)
  • Static analysis with capability inference
  • Trust scoring with 4-signal algebra (L0 through L3)
  • Lockfile generation (skill-lock.json)
  • CycloneDX 1.6 ASBOM output
  • 540-skill benchmark (96.95% F1, 100% precision)
  • 5 CLI commands: scan, verify, lock, trust, sbom
  • Support for 3 formats: Claude Code, MCP, OpenClaw
  • 31-page research paper with 5 theorems

v0.2.0 -- Expanded Coverage (February 2026)

Broadened format support and improved detection.

  • 3 additional agent framework parsers
  • Expanded benchmark with community samples
  • CI/CD integration improvements
  • Refactored codebase to enterprise-grade modularity

v0.3.0 -- System Discovery and Dashboard (March 2026)

Transformed SkillFortify from a project scanner into a system-wide security tool.

  • 22 agent frameworks supported (up from 6)
  • System auto-discovery: skillfortify scan with no arguments scans all AI tools
  • 23 IDE profiles: Claude Code, Cursor, VS Code, Windsurf, Gemini CLI, Cline, Continue, GitHub Copilot, n8n, Roo Code, Trae, Kiro, Kode, Jules, Junie, Codex CLI, and more
  • HTML Dashboard: skillfortify dashboard generates standalone visual security reports
  • Framework listing: skillfortify frameworks displays all supported frameworks
  • Registry scanning: skillfortify registry-scan for MCP, PyPI, and npm marketplaces
  • 1,818 tests (up from 675)
  • 9 CLI commands (up from 5)

Planned

v0.4 -- Advanced Detection

  • Install-time attack detection: Typosquatting, dependency confusion, namespace squatting
  • Enhanced registry scanning: Deeper analysis of remote marketplace entries
  • Policy engine: Define organizational rules for skill approval (minimum trust levels, required capabilities, blocked patterns)

v0.5 -- Runtime and IDE Integration

  • Runtime monitoring: Capability enforcement during skill execution, not just at scan time
  • VS Code extension: Inline verification as you edit skill files
  • Official GitHub Action: skillfortify/scan@v1 with pull request annotations

v1.0 -- Enterprise Features (Vision)

  • Fleet dashboard: Centralized visibility across all agent projects in an organization
  • Cryptographic skill signing: Keyless signing protocol for skill authors
  • Advanced composition analysis: Cross-skill interaction rules and policy enforcement
  • Behavioral contract integration: Skills verified against agent-level specifications

How to Contribute

SkillFortify is open source (MIT License). Contributions are welcome:

  • New framework parsers: Add detection for frameworks not yet covered
  • Benchmark skills: Contribute samples to SkillFortifyBench
  • Bug reports: github.com/varun369/skillfortify/issues
  • Documentation: Improve wiki pages, add tutorials
  • Research: Extend the formal model, propose new theorems

See CONTRIBUTING.md for guidelines.

Feature Requests

Open a discussion on GitHub Discussions or file a feature request as an issue. The roadmap is shaped by real developer needs.

SkillFortify -- formal foundations for trustworthy AI agents.

Clone this wiki locally