devflow-kit 3.2.0 → 3.3.0

package/src/assets/skills/worktree-support/references/discovery.md

@@ -0,0 +1,48 @@
+# Worktree Discovery
+
+How commands that auto-discover worktrees (`/code-review`, `/resolve`) find them, and how the count found sets the flow. The protected-branch list lives only in the skill's `## Protected Branches (Canonical List)`; this file points to it and copies none of it.
+
+## Algorithm
+
+### Step 1: List Worktrees
+
+Run `git worktree list --porcelain`
+
+### Step 2: Parse Each Entry
+
+Extract: `worktree {path}`, `HEAD {sha}`, `branch refs/heads/{name}`
+
+### Step 3: Filter
+
+Exclude:
+- **Bare worktrees** (no branch)
+- **Detached HEAD** (no named branch; roots and detached HEAD: `roots.md`)
+- **Protected branches**: every branch in the skill's `## Protected Branches (Canonical List)`
+- **Mid-rebase or mid-merge**: check `git -C {path} status` for "rebase in progress" or "merging"
+
+### Step 4: Check for Work
+
+Each worktree must have commits ahead of base branch. Skip those that are clean/up-to-date.
+
+### Step 5: Deduplicate
+
+If two worktrees are on the same branch, use only the first worktree's path.
+
+### Step 6: Sort
+
+Sort by last commit date (most recent first).
+
+### Step 7: Return
+
+Return list of reviewable worktrees with path, branch, HEAD SHA.
+
+## Multi-Worktree Mode
+
+| Condition | Behavior |
+|-----------|----------|
+| 0 reviewable worktrees | Error: no reviewable branches found |
+| 1 worktree (common case) | Single-worktree flow, zero behavior change |
+| 2+ worktrees | Report count, process worktrees **sequentially** (one at a time) |
+| 5+ worktrees | Report count and proceed — user manages their worktree count |
+| `--path` flag provided | Use only that worktree, skip discovery |
+| Same branch in 2 worktrees | Deduplicate — review once using first worktree's path |

package/src/assets/skills/worktree-support/references/roots.md

@@ -1,6 +1,6 @@
 # Project Roots and Detached HEAD
 
-Where devflow keeps per-project data across checkouts, and what it does on a detached HEAD. Companion to the worktree discovery algorithm in `SKILL.md`.
+Where devflow keeps per-project data across checkouts, and what it does on a detached HEAD. Companion to the worktree discovery algorithm in `discovery.md`.
 
 ## Which Root Holds What
 
@@ -24,6 +24,6 @@ A detached HEAD is a commit checked out without a branch (`git checkout <sha>`,
 | Surface | Behaviour |
 |---------|-----------|
 | Knowledge commit | Skipped: a commit there becomes unreachable once HEAD moves. The Knowledge agent reports `KB_COMMIT: skipped (detached HEAD) — uncommitted: <paths>`, and the calling workflow names those paths to the user |
-| Worktree discovery | Excluded from auto-review (discovery Step 3 in `SKILL.md`): there is no branch to review, compare or push |
+| Worktree discovery | Excluded from auto-review (discovery Step 3 in `discovery.md`): there is no branch to review, compare or push |
 | Memory bootstrap | The first compaction bootstraps `WORKING-MEMORY.md` stamped `branch: (detached)`, keyed on the HEAD commit; an unborn branch has no commit and still skips |
 | Memory header | Renders `detached @ <short-sha>` instead of `on unknown` |