AI agent orchestration for iterative spec-driven development
deno run -A jsr:@cdaringe/ralphmania -i 10A task runner that thinks for itself — and checks its own work.
Runs AI agents in a loop over your specification, making progress on each scenario until completion.
Each scenario is verified before being marked done. Failed validation triggers targeted rework.
Configurable escalation strategies let you widen search or retry with different agents on failure.
Plugin hooks let you intercept every stage — before, during, and after each scenario run.
Five steps from spec to shipped.
Author a markdown table of scenarios in specification.md.
Execute deno run -A jsr:@cdaringe/ralphmania -i 10 to start the loop.
An AI agent is dispatched per scenario, running in an isolated git worktree.
Completed scenarios are validated. Failures are flagged NEEDS_REWORK and retried.
A living progress.md file records every scenario's status throughout the run.