Amaleh — AI skill for Claude Code

Amaleh

**عمله** is pronounced **ʿamaleh** (roughly **Ah-mah-leh**) and means **workers / laborers** in Persian. *Amaleh* is the project's Latin-script name: coordinated workers contributing to a shared, verified outcome.

**Why Amaleh exists: top-tier models are too expensive to do the work.** Without structure, an expensive model spends its premium context writing every edit, every check, every micro-decision. Amaleh turns that model into a director instead of a laborer: it segments work into chunks with clear outcomes, delegates each chunk to cheap Flash models, and only verifies direction at chunk boundaries. The result is the same verified delivery at a fraction of the token cost.

A self-contained workflow skill for Codex and Claude: discovery, planning, delegated parallel implementation, independent review and verified delivery. Disk checkpoints let another session resume the work.

How the division of labor works:

• Coordinator (expensive model, minimal turns): clarifies intent, chunks the work, defines acceptance criteria, integrates results. One delegate call per chunk; no per-step instructions, no hand-written briefs, no micro-management.
• Workers (cheap Flash models via pi/OpenRouter): own a chunk end to end — implementation, checks, repair cycles. When uncertain, they consult Jev directly through a bundled helper instead of escalating to the coordinator.
• Reviewer (the other Flash family, read-only): independently verifies each chunk with structured coverage; findings route back into the worker's repair loop, not to the coordinator.
• Jev (cheap decision model): answers bounded either/or questions for workers and the coordinator; TypeScript code — not any model — enforces dependencies, ownership, checks and review coverage.
• Routing: deterministic round-robin across eligible model families seeded by the run's session hash, so work spreads across vendors instead of fixating on one.
• Escalation: only genuine boundaries