| name | pipeline-renderer |
| description | Dual-Stage Decoupled Pipeline for natural, complete, uninhibited narrative and scenario expansion without placeholder token degeneration. |
Dual-Stage Pipeline Renderer Skill
This skill provides an automated two-stage decoupled generation workflow to eliminate robotic placeholder tokens (ROLE_A, ROLE_B) and meta-framing disclaimers from output.
Architecture
- Stage 1 (Planner): Extracts an outline
SceneBlueprint with real character names, environmental settings, and progressive plot beats.
- Stage 2 (Renderer): Unpacks the blueprint into high-fidelity prose, allocating 100% of model attention to narrative depth, dialogue, and pacing.
Usage
Run via CLI:
python .agents/tools/re-toolkit/cli.py pipeline "Your complex scenario or narrative request"
Output full blueprint & prose in JSON:
python .agents/tools/re-toolkit/cli.py pipeline "Your request" --json
Use in Python:
from pipeline import PipelineEngine
engine = PipelineEngine()
result = engine.execute("Describe a dramatic negotiation between two rival founders.")
print(result["rendered_text"])
