100%

06 — runtime-skill-tests

Goal

Land the script-first skill + runtime + tests: thin .claude/skills/consolidate/SKILL.md (role + @-includes + action-consumption table only) over scripts/consolidate_runtime/ (gates/scope/candidates/dispatch/api + next_action.schema.json + tests/). The skill body carries NO orchestration algorithm — the runtime owns gate eval, scope assembly, freeze, and the executor loop; the skill consumes structured next_action.

Implementation Sketch

Author scripts/consolidate_runtime/ per script-first §4 (api.py public surface, gates/scope/candidates/dispatch modules, next_action.schema.json, tests/). Author the thin SKILL.md (role: hybrid; @-include routing/skill-control-contract/script-first/state-discipline; consult §N annotations; action-consumption table). Full pytest coverage: gate, scope, freeze, judge-dispatch shape, executor-loop, —self-test.

Spec References

.claude/rules/script-first.md §1/§4/§8; .claude/rules/skill-template.md (thin body); .claude/rules/workflow-ownership.md §1 (scripts drive the loop)

Work Items

  • scripts/consolidate_runtime/ per §4 layout (api/gates/scope/candidates/dispatch + schema + tests)
  • Thin .claude/skills/consolidate/SKILL.md (role + @-includes + action-consumption only)
  • next_action.schema.json + --self-test end-to-end
  • pytest: gates/scope/candidates/dispatch + script-first audit clean + prose-lint clean