OmniNode-ai

/onex:coverage_sweep — one command, one typed result

Measure test coverage across all Python repos under the workspace, flag modules below threshold,

OmniNode-ai 4 3 Actualitzat fa 3 mesos

Recursos

2
GitHub

Instal·la

npx skillscat add omninode-ai/omniclaude/plugins-onex-skills-coverage-sweep

Instal·la des del registre de SkillsCat.

SKILL.md

/onex:coverage_sweep — one command, one typed result

Skill ID: onex:coverage_sweep · Command: uv run onex skill coverage_sweep (omnibase_infra) · Backing node: node_coverage_sweep (omnimarket)

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[CoverageSweepResult] 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.

What this skill does NOT do

  • Construct a payload file, cd anywhere, or cat a workflow_result.json (all internal to onex 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.nodes.node_coverage_sweep.handlers.handler_coverage_sweep.CoverageSweepResult