Skip to content

v1.7.24

Choose a tag to compare

View all tags
@Alan-TheGentleman Alan-TheGentleman released this 25 Mar 11:21
v1.7.24
bd12598

v1.7.24 — Self-Sufficient Sub-Agents + Upgrade UX + Per-Agent Model Routing

Highlights

Sub-agents now discover skills on their own — SDD phase sub-agents no longer depend on the orchestrator to provide skill paths. Each sub-agent self-discovers the skill registry via engram or .atl/skill-registry.md, then loads relevant skills for the current task. A sdd-apply on a React project loads React patterns; on Go it loads Go testing conventions — automatically.

Per-agent model routing in OpenCode — The background-agents plugin now passes each agent's configured model through session.prompt(). Configure sdd-spec with Opus and sdd-explore with Sonnet, and each phase runs on its assigned model.

Upgrade command shows progress — gentle-ai upgrade now shows real-time status instead of a blank screen.

Documentation — New "Sub-Agents — Smarter Than You Think" section in docs/intended-usage.md.

All changes since v1.7.22

  • feat(sdd): self-sufficient sub-agents with skill-registry auto-discovery
  • feat(opencode): pass agent model through delegate() via session.prompt()
  • feat(upgrade): add progress output during upgrade execution
  • docs: intended-usage guide, SDD/engram docs, skill-registry usage

Full changelog: v1.7.22...v1.7.24

Assets 9
Loading