git-worktrees: rename to match its provisioned-parallel-dev identity (candidate: parallel-dev; decide the worktree-cleanup pairing) #64

New issue

Open

opened 2026-08-25 09:04:01 +00:00 by jbr870 · 0 comments

jbr870 commented

2026-08-25 09:04:01 +00:00

Owner

Problem

After #63, git-worktrees explicitly disowns bare worktrees — its identity is provisioned
parallel development (the parallel_dev: slot machinery: worktree + ports + database +
per-slot container axes, plus the hooks-based ad-hoc path). The name now describes the one case
the skill routes away, and a name is the strongest routing signal a skill has: "create worktree"
next to a skill named git-worktrees pulls against the new "defer to the harness's native
worktree flow" branch.

Proposal

Rename to match the declaration the skill consumes — leading candidate parallel-dev
(mirrors parallel_dev:, same precedent as the adapter skills being named after
forge.adapter: values). Decisions the rename must make, not dodge:

The pair: slot retire routes to worktree-cleanup; renaming one side leaves the pair
incoherent. Rename it too (e.g. alongside the slot vocabulary), fold it, or keep it — decide.
Trigger vocabulary stays human: users say "worktree"/"slot"; triggers keep those words even
as the skill name moves.
Reference surface (measured 2026-08-25): ~13 refs in 5 shipped files
(slot-isolation.md ×7, worktree-cleanup ×3, develop, git-worktrees itself, a slot test
fixture), README name table, repo CLAUDE.md, scripts/lint-conventions.sh,
scripts/rewrite-plugin-refs.sh, directory/frontmatter identity (plugin validate), and
consumer-project guidance on both machines.

Gate on priority

Deliberately parked until #63's routing guards are validated by the first parallel-session wave
on devwork-skills. If no session misroutes, this is cosmetic — batch it with other work. If a
session reaches for this skill despite the guards, that occurrence is the evidence that raises
this ticket's priority.

Lane

Full pipeline (invocation-surface rename across shipped files + consumer-facing docs + release
notes) — not the meta lane.

## Problem After #63, `git-worktrees` explicitly disowns bare worktrees — its identity is **provisioned parallel development** (the `parallel_dev:` slot machinery: worktree + ports + database + per-slot container axes, plus the hooks-based ad-hoc path). The name now describes the one case the skill routes away, and a name is the strongest routing signal a skill has: "create worktree" next to a skill *named* git-worktrees pulls against the new "defer to the harness's native worktree flow" branch. ## Proposal Rename to match the declaration the skill consumes — leading candidate **`parallel-dev`** (mirrors `parallel_dev:`, same precedent as the adapter skills being named after `forge.adapter:` values). Decisions the rename must make, not dodge: - **The pair:** slot retire routes to `worktree-cleanup`; renaming one side leaves the pair incoherent. Rename it too (e.g. alongside the slot vocabulary), fold it, or keep it — decide. - **Trigger vocabulary stays human:** users say "worktree"/"slot"; triggers keep those words even as the skill name moves. - **Reference surface** (measured 2026-08-25): ~13 refs in 5 shipped files (slot-isolation.md ×7, worktree-cleanup ×3, develop, git-worktrees itself, a slot test fixture), README name table, repo CLAUDE.md, `scripts/lint-conventions.sh`, `scripts/rewrite-plugin-refs.sh`, directory/frontmatter identity (plugin validate), and consumer-project guidance on both machines. ## Gate on priority Deliberately parked until #63's routing guards are validated by the first parallel-session wave on devwork-skills. If no session misroutes, this is cosmetic — batch it with other work. If a session reaches for this skill despite the guards, that occurrence is the evidence that raises this ticket's priority. ## Lane Full pipeline (invocation-surface rename across shipped files + consumer-facing docs + release notes) — not the meta lane.

jbr870 added the

labels

2026-08-25 09:04:01 +00:00