Production Recipes

These recipes ship with amplihack and demonstrate real-world workflow patterns at scale.

Source: amplifier-bundle/recipes/

Development Workflows

Recipe	Description
default-workflow	Complete development lifecycle: requirements → design → implement → test → merge
verification-workflow	Lightweight workflow for trivial changes: config edits, doc updates, single-file fixes
qa-workflow	Minimal workflow for simple questions and informational requests
investigation-workflow	Systematic investigation with parallel agent deployment
guide	Interactive guide to amplihack features

Quality & Reliability

Recipe	Description
quality-audit-cycle	Iterative audit loop: lint → analyze → fix → re-lint → verify improvement
self-improvement-loop	Closed-loop eval improvement: eval → analyze → research → improve → re-eval → compare
domain-agent-eval	Evaluate domain agents: eval harness + teaching evaluation + combined report
long-horizon-memory-eval	1000-turn memory stress test with self-improvement loop
sdk-comparison	Run L1-L12 eval on all 4 SDKs and generate comparative report

Multi-Agent Decision Making

Recipe	Description
consensus-workflow	Multi-agent consensus at critical decision points with structured checkpoints
debate-workflow	Multi-agent structured debate for complex decisions requiring diverse perspectives
n-version-workflow	N-version programming: generate multiple independent implementations, pick best
cascade-workflow	3-level fallback cascade: primary → secondary → tertiary

Orchestration

Recipe	Description
smart-orchestrator	Task classifier + goal-seeking loop with up to 3 execution rounds
auto-workflow	Autonomous multi-turn workflow — continues until task complete or max iterations

Migration

Recipe	Description
oxidizer-workflow	Automated Python-to-Rust migration with quality audit cycles and degradation checks