cell-divide

To clone a skill genome, you copy the root .Codex/skills/ contents into a target cell subtree and update each child SKILL.md with the parent commit SHA in the YAML lineage. This ensures auditable evolution and tracks repo replication accurately across cells.

Skill lineage tracking records the parent commit SHA and origin metadata for each cloned skill cell during decomposition workflows. You need it when new cells must be auditable, registered, and traceable back to their original repository root genome for later execution.

You register a new skill cell by inserting a row into lattice/knowledge/skills_registry describing its location, content, and lineage origin. This registration occurs alongside a separate reviewed git commit to maintain strict path validation and auditable evolution.

Cell-divide exhibits phase-safe inert behavior during early rollout and does not execute recursive division until later phases. This reduces risk by limiting Phase 1 to auditing, path validation within meta/harness/sections/*, and controlled skill-cell creation only.

You normalize YAML lineage and local adaptations by applying strict path validation within meta/harness/sections/* and updating child SKILL.md files with the parent_commit_sha. This standardizes metadata formatting during controlled skill-cell creation and registry insertion.

No, cell-divide requires strict path validation within meta/harness/sections/* for target paths. This constraint ensures controlled skill-cell creation during decomposition workflows, maintaining auditable lineage and proper registration in the skills registry.