feat: add streaming vuln+exploit pipeline and per-agent model selection

[TedWei]

• Implement streaming execution where each exploit starts immediately when its corresponding vuln agent completes, reducing total execution time
• Add per-agent model mapping (Haiku for simpler tasks like xss/ssrf, Sonnet for complex tasks like injection/auth analysis)
• Add partial completion recovery for pre-recon agent to preserve intermediate results when final report generation fails
• Add Docker support with docker-compose.yml and helper scripts
• Update .gitignore for environment files and session data

🤖 Generated with Claude Code

[keygraphVarun]

Hi @TedWei
My sincere apologies for the late reply. This slipped through the cracks and that's on me.
Thanks for the thought you put into this. Unfortunately we're not accepting external PRs at this stage (noted in our README). We're a small team iterating fast on core architecture, which makes coordinating outside contributions tricky. We may open up contributions down the line once we sort out CLA and other logistics.
Since you opened this, we've also done major refactors (TypeScript, Temporal) so a lot of what you touched has changed significantly.
If there are features you'd still like to see, please open an issue or discussion thread. Happy to consider them as we plan future work.
Appreciate the effort, and sorry again for leaving you hanging.