Skip to content

v5.13.0 — AG Wave D persona roster

Latest

Choose a tag to compare

@alexherrero alexherrero released this 07 Jul 01:59
f554232

MINOR. The persona roster from agentm-personas.md goes from 2 authored manifests to 11: the four activation axes (tier:/opinions:/modes:/triggers:) are retrofitted onto brain and team-coordinator, and the 9 remaining designed-only rows (Architect, Designer, Tech-Lead, Engineer, Reviewer, Operator, Troubleshooter/SRE, Researcher, Maintainer) are authored for the first time. A fresh-session re-audit confirms agent-def effort: frontmatter is a real, currently-documented dispatch-time binding — closing the P12 Task 2b open question. content-refresh (already shipped in crickets) gets its first named consumer: a re-pin entrypoint for the model-effort-routing chart's pinned model-id strings.

Added

  • 9 new kind: persona manifests (personas/{architect,designer,tech-lead,engineer,reviewer,operator,troubleshooter,researcher,maintainer}.md).
  • scripts/model_effort_routing_refresh.py — content-refresh's re-pin entrypoint for the tier-scale chart's model-id strings.

Changed

  • brain.md and team-coordinator.md gain the four activation axes.
  • agentm-model-effort-routing.md's stale doc-drift citation corrected + a new amendment recording the fresh-session effort:-binding re-audit outcome (confirmed BINDS).

Internal

  • Test coverage extended in test_check_personas.py, test_persona_resolve.py, and a new test_model_effort_routing_refresh.py.
  • Workflow-step automatic-adoption (plan task 3) deferred, not shipped — a genuine cross-repo design-contract blocker (crickets' phase-command prose is the source of truth for a workflow step, per agentm-persona-activation.md's own locked design call); reported rather than forced through.

Full CHANGELOG entry: CHANGELOG.md#v5130

PR: #239