ak-watzup

name	ak:watzup
description	Generate short handoff reports from Git branches, remote refs, worktrees, unfinished plans, and roadmap docs. Surfaces priority-ranked next steps with checkbox progress and rationale. Use when the user asks what's in flight, wants progress/next steps, is in a fresh worktree or detached checkout, or needs end-of-session status.
user-invocable	true
when_to_use	Invoke for end-of-session handoffs, progress summaries, cross-branch worktree status, unfinished plan discovery, and next-step recommendations.
category	utilities
keywords	["session","wrap-up","changes","review","worktree","branches","plans","roadmap","priority","next-steps"]
metadata	{"author":"agentkit","version":"1.2.0"}

Wrap Up

Create a short, evidence-backed handoff report for the active project, with priority-ranked next steps grounded in plan progress and roadmap state.

This skill handles status and handoff reporting only. It does not implement, edit, commit, checkout, merge, push, or fetch unless the user explicitly requests fresh remote refs.

For conversation-state compaction that preserves decisions, rationale, and session context for a fresh agent, use ak-handoff instead.

Required Scan

Run the scanner first from the project root:

node scripts/watzup-scan.cjs --json

Use --fetch only when the user asks to refresh remotes before the report:

node scripts/watzup-scan.cjs --json --fetch

When developing from this source repository before install, run the same command from the source skill directory.

Default behavior:

• Scan local branches and remote branch refs.
• Scan registered worktrees.
• Scan unfinished plans from visible worktrees and tracked branch refs.
• Count - [ ] / - [x] checkboxes in each plan directory (plan.md + phase-*.md) for progress %.
• Scan docs/*roadmap*.md and docs/*milestones*.md for active milestones.
• Build priority-ranked next steps via composite scoring (see below).
• Do not run network operations.
• Do not change branches or mutate the checkout.

Priority Ranking

The scanner now emits nextSteps[] as objects with {priority, action, rationale, planId?}. Ordering reflects a composite score per plan:

• Status: in-progress (+400) > in-review (+300) > pending (+150).
• Workspace alignment: current worktree (+600), current branch (+400).
• Provenance: filesystem source (+80), local ref (+40).
• Momentum: plans between 40-90% complete get bumped (close to done). Brand-new plans (<10%) get a small starter bump.

Hygiene steps (dirty working tree, detached HEAD) always rank first. Roadmap milestones fill remaining slots after plan-driven actions.

Report Format

Keep output brief. Prefer this structure:

1. Current State - branch or detached HEAD, dirty/clean, active worktree.
2. Recent Work - only the highest-signal branches/worktrees.
3. In-Flight Plans - unfinished plans with X/Y todos · NN% done annotation.
4. Roadmaps - active milestones from docs/*roadmap*.md, if any.
5. Next Steps - 5 to 6 priority-ranked actions, each with one-line rationale.
6. Warnings - scanner failures, stale remote-ref caveat, detached HEAD.

If the scanner fails, say it failed and include the error. Then use minimal read-only fallback commands:

git status --short --branch
git worktree list --porcelain
git for-each-ref --format='%(refname:short) %(committerdate:iso8601) %(objectname:short) %(subject)' refs/heads refs/remotes
find plans -maxdepth 2 -name plan.md -print
find docs -maxdepth 2 -iname '*roadmap*.md' -print

Do not pretend the full scan completed when fallback was used.