OpenCode Agent Loop v0.1.0 — Now Public!

[wryan2986]

After months of internal use, the OpenCode Agent
Loop is now open source under MIT.
What it does: coordinates planning, implementation,
testing, review, and escalation
agents through a single /feature
command inside OpenCode.
Key features:
- 10 specialized agent roles with configurable
model routing
- Free-first with paid fallback across 5+ providers
- Safety-first: read-only reviewers, no-git-push
for workers, recursion guards
- Stage-enforced workflow: plan → approve → test →
build → verify → review → commit
- Provider-wide cooldown management for rate limits
- Privacy-aware routing with data-policy
classification
- Local Qwythos 9B model support for sensitive/
offline tasks
Requires a patched OpenCode for subagent failure
exposure (patch included).
https://github.com/wryan2986/opencode-agent-loop
Would love feedback, issues, and contributions!