Skip to content

Agent Backbone 0.4.4

Choose a tag to compare

@eandualem eandualem released this 30 Oct 19:55
b3f36e2

Agent Backbone 0.4.4 adds clearer operating guidance for agents and preserves existing installation choices.

  • Fresh core-v1 setups now include starter-swarms, explaining when an agent should use a swarm, how to choose specialist help, and how the initiating agent retains responsibility. Detailed swarm role policies remain role-specific.
  • Starter learning and collaboration guidance makes specialist expertise and one continuing repository/task owner explicit.
  • Current plans now describe one concrete current task as its real steps. Blocked steps name who or what they need and the required result; task changes use a new plan ID while preserving unfinished work in the agent's chosen continuity records.
  • Existing five-policy installations remain compatible with first-agent setup. Upgrades preserve editable policy copies and assignments; they do not automatically replace custom instructions or assign the new starter policy.

The setup schema, agent-state API and Current plan API are unchanged. No structured dependency field is introduced.

Validation: 4,874 tests, lint/format, independent feature and release reviews, and private packaged fresh-install/upgrade checks. Private checks verified completed setup replay, preserved edits and credentials, and authenticated socket/API access without launching agents or installing a user service.

Upgrade notes: finish pending setup with its saved provider before upgrading. Existing Linux service migration precautions from 0.4.3 still apply; this release does not change service lifecycle behavior.

Changes: #539 and #540.