| name | agentic-reactor |
| description | Overarching patterns for self-guarded, pause-aware, agent-driven desktop/web platforms. Combines finder-reactor, Tauri shell, X resources, and CV guard into a cohesive autonomous system with intervention only on guards. Use for high-level autonomy design, meta-improvement loops, MCP composability. Fusion for the reactor as living system; fission for specific guards. |
Agentic Reactor — The Living, Self-Improving Finder Platform
Core Mission: Make an opportunity-finder not just a tool, but an autonomous agent that runs with minimal human input, using self-guards and pauses to stay safe. It decides, acts (search, prep), learns, and improves — surfacing only when necessary (low fit, high cost, CV changes).
Principles
- Autonomy with Guardrails: Reactor loop runs search → analyze → decide → prep if cleared → track.
- Pauses: Explicit intervention points via UI dialogs and MCP pause responses.
- Smart Decisions: Structured "next action" with guards evaluated in the core (Rust or backend service).
- Self-Improvement: After cycles, surplus proposals — auto-apply low-risk, pause for high-impact.
- Composability: Full MCP exposure so Cursor, Grok, or other agents can invoke guarded hunts.
- Integration: X via x-agent-resources; CV via cv-promote-guard; shell via tauri-agentic.
Architecture (Fusion View)
FinderReactor (core):
- State: leads/preps in app data (JSON or SQLite).
- Guards: Cost, Rate, Fit, CVPromote (delegates to cv-promote-guard).
- X layer: x-agent-resources patterns.
- MCP server: reactor methods +
ask_user for pauses.
- Optional background daemon with tray/UI surfacing pauses.
Shell (Tauri or web):
- Command palette + guard dashboards + pause modals.
- Frontend: react-client-expert minimal state.
Surplus Protocol
After reactor-related work, document Q-factor wins and suggested follow-ups. Track in surplus-log or fusion-state.
Activation
High-level design: load agentic-reactor + finder-reactor + tauri-agentic + fusion-sage.
Example provenance: collab-finder.
