horizon-offload

Featured

Offload repository development issues to a persistent Horizon Linux worker, using repository-defined environments and a configured Azure or local Docker profile. Inspect and reconnect to existing tasks without replay.

AI & Automation 710 stars 28 forks Updated today MIT

Install

View on GitHub

Quality Score: 92/100

Stars 20%
95
Recency 20%
100
Frontmatter 20%
70
Documentation 15%
100
Issue Health 10%
50
License 10%
100
Description 5%
100

Skill Content

# Horizon issue offload Use `horizon-worker` for the persistent allocation, Git handoff, saved task start and observation. Discover the controller, `gh`, and tooling for the selected provider before asking the user to install anything. The first agent lane uses Codex CLI. Read the repository's instructions and issue acceptance criteria before constructing the task. ## Repository environments Look for `.horizon/worker.yml` at the exact selected Git commit. With no local checkout, read it through GitHub's contents API using that ref; a local checkout is not required. Validate a temporary copy with `horizon-worker manifest <file>`. The version-1 shape is: ```yaml version: 1 default_environment: frontend environments: frontend: image: registry.example/web-worker@sha256:<64 lowercase hex digits> directory: apps/web checks: test: [npm, test] backend: image: registry.example/api-worker@sha256:<64 lowercase hex digits> directory: services/api checks: test: [cargo, test] ``` Each environment is one worker image. Select an explicitly requested environment, otherwise the declared default or the sole entry. For ambiguous multi-environment issues, inspect affected paths and clarify the selection if needed. Cross-cutting issues can use separate task receipts per environment; this does not orchestrate networked service containers. An absent manifest calls for a proposed starter configuration, not an invented registry or mutable `latest` image. C...

Details

Author
peters
Repository
peters/horizon
Created
6 months ago
Last Updated
today
Language
Rust
License
MIT

Integrates with

Similar Skills

Semantically similar based on skill content — not just same category