Content Review

The Reviewer Agent checks content safety and quality before publishing.

Review Dimensions

1. Sensitive Words — Configurable word list, no built-in defaults
2. Platform Compliance — Title/content length, forbidden patterns
3. Quality Score — 0-100 across 9 criteria

Pipeline

Writer → Reviewer → Publisher
              ↓ failed
            Skip publish + error

Configuration

reviewer = ReviewerAgent(llm_client, config={
    "reviewer_sensitive_words": {"spam": "spam detected"},
    "reviewer_min_score": 60,
})

Pipeline Usage

task = orchestrator.create_task(
    task_type="write_review_publish",
    params={"topic": "...", "platform": "juejin"}
)
result = orchestrator.execute_pipeline(task)