OmniNode-ai /onex:pr_review — one command, one typed result
Skill ID: onex:pr_review · Command: uv run onex skill pr_review (omnibase_infra) · Backing node: node_skill_pr_review_orchestrator (omniclaude)
A dispatch skill IS one CLI call. Payload construction, node dispatch, and
result extraction all live in the onex skill entrypoint (declarative
skill_mapping.yaml registry) — there is no procedure to learn here. The
command prints exactly one typed ModelSkillResult[ReviewVerdict] JSON to
stdout carrying the FULL handler result; RuntimeLocal logs and intermediate
context go to a capture file + the artifact store, never to you.
See prompt.md for the one command and how to present the typed result.
Routing Contract
The uv run onex skill pr_review entrypoint publishes to onex.cmd.omnimarket.pr-review-bot-start.v1
(the node_pr_review_orchestrator start topic) through receipt-mode dispatch. If routing fails,
surface SkillRoutingError directly; do not produce prose.
What this skill does NOT do
- Construct a payload file,
cdanywhere, orcata workflow_result.json (all internal toonex skill) - Run any inline scan, probe, or orchestration — the backing node owns all logic
- Contain executable logic in this directory — markdown only
Related
- CLI entrypoint:
omnibase_infra/src/omnibase_infra/cli/cli_skill.py - Skill→node mapping:
omnibase_infra/src/omnibase_infra/cli/skill_mapping.yaml - Result model:
omnimarket.review.pr_review_io.ReviewVerdict