Adopt beads (bd) for AI task tracking

[J-MaFf]

Adopt beads (bd, Dolt-backed) as the dependency-graph task/memory layer beneath GitHub Issues, following the pattern trialed in J-MaFf/winget-app-setup#148.

Layered model

• GitHub Issue = the shippable unit (branch → PR → Fixes #N → squash-merge).
• Beads = the execution/memory layer underneath — the bd ready queue, a dependency graph, and bd remember persistent memory.

Steps (full workflow: the git-policies skill's .beads/ section)

• bd init (embedded Dolt). Wire the Dolt remote to origin if not auto-set: bd dolt remote add origin git+https://github.com/J-MaFf/python_functions.git
• bd setup claude (SessionStart/PreCompact bd prime hooks)
• Reconcile the bd init CLAUDE.md block with git-policies — keep bd's queue / bd remember / durable-sync, but session-close pushes the feature branch + bd dolt push and the merge stays human-gated; scope the memory rule to repo-level. ⚠️ Don't run bd setup claude afterward to clear the "stale" warning (it reverts the reconciliation).
• Confirm the .beads/ commit convention (track issues.jsonl + config.yaml, ignore embeddeddolt/)
• bd dolt push; verify refs/dolt/data on origin (git ls-remote origin 'refs/dolt/*')
• Update STATUS.md / CHANGELOG.md per git-policies

Reference: trial J-MaFf/winget-app-setup#148 · workflow J-MaFf/claude-skills#93 · ClickUp 86baggbgc