{% extends "base.html" %} {% block content %}
Forja
Preparation
Execution Pipeline
Human Oversight & Completion
PHASE 1 — PREPARATION
Definition & Guardrails
1. Agent Definition
Versioned YAML with prompt, model, output schema and risk threshold
2. Policy Application
Guardrails policy loaded and attached to the agent
PHASE 2 — EXECUTION PIPELINE
The Main Processing Line
3. Input Guardrails
Validates input against policy before it reaches the LLM
4. LLM Reasoning
The agent generates structured output using the configured model
5. Output Guardrails
Validates the LLM response for safety and correctness
6. Action Proposal
Generates actions with risk scores and rollback plans
PHASE 3 — HUMAN OVERSIGHT & COMPLETION
Critical Gate + Finalization
{% endblock %}