← ClaudeAtlas

airflow-software-factorylisted

Operate and extend an Apache Airflow software factory for coding agents: turn GitHub issues and work orders into tested, reviewed pull requests with human approval gates, durable Factory Cells, sandbox isolation, bounded work graphs, recovery, evidence and provenance, CI/CD, and Rust/Python operator surfaces. Use when adopting, configuring, running, debugging, deploying, auditing, or improving swfactory, factory.toml, Airflow blueprints, GitHub publication, agent sandboxes, or agentic software-delivery workflows. Keep Airflow as the sole lifecycle scheduler.
zozo123/ariflow-swfactory · ★ 1 · AI & Automation · score 72
Install: claude install-skill zozo123/ariflow-swfactory
# Airflow software factory Build a repeatable software production system on Apache Airflow. Treat the GitHub issue as a work order, the blueprint as a production route, Airflow as the plant scheduler, and each sandbox as one work cell. Airflow expands issue-to-target jobs, pauses for human decisions, retries infrastructure failures, records state, and exposes operations. A coding agent performs the bounded job. Git keeps the trace record. A human keeps the merge key. ```text WORK ORDER -> ROUTE -> INTENT GATE -> SPEC -> PLAN GATE -> WORK CELL -> TESTS + REVIEW -> VALIDATED PATCH -> PULL REQUEST -> HUMAN MERGE | v CONTINUOUS IMPROVEMENT ``` Install this skill from the public repository with Vercel's open `skills` CLI: ```bash npx skills add zozo123/ariflow-swfactory --skill airflow-software-factory ``` Use it without installing: ```bash npx skills use zozo123/ariflow-swfactory@airflow-software-factory ``` Browse the public listing at `https://skills.sh/zozo123/ariflow-swfactory/airflow-software-factory`. ## Choose the job Infer the narrowest useful mode from the request: - **Demo:** run the scripted, keyless route to inspect behavior without model calls. - **Adopt:** add `factory.toml`, select a blueprint, and connect one target repository. - **Operate:** trigger, approve, inspec