Parallelize work on cheap models
cheap-fanout-synthesisworkflowsetup L2★0
sashabogi/token-scrooge ↗What it does
Low-budget parallel research synthesis without losing findings
Best for
Fan-out cheap research when no single synthesizer can hold the full pool.
Inputs
- · args (JSON configuration)
- · test probes or findings array
- · product brief
- · file paths or repo root
Outputs
- · structured report with citations
Requires
- · agent()
- · parallel()
Preconditions
- · Agent runtime available (Claude Code Workflow)
Failure modes
- · Subagent failures silently dropped, partial results returned
Trust signals
- · Deterministic orchestration with phase gates