zskills-dev

Safe commit workflow with optional scope hint. Inventories all changes, classifies related vs. unrelated files, traces dependencies, protects other agents' work, and optionally pushes, lands worktree commits, or opens a PR via /land-pr. Positional auto enables auto-merge (PR mode only).

Lightweight task dispatcher for ad-hoc work: documentation, examples, refactoring, content updates. Worktree/direct/pr landing modes via flag or execution.landing config. Recurring via every SCHEDULE; stop/next manage the schedule.

Draft a high-quality plan through iterative adversarial review: research, draft, review, devil's-advocate, refine — repeated until convergence. Output is a plan file ready for /run-plan execution.

Draft test specifications into an existing plan through iterative adversarial review. Appends a `### Tests` subsection per pending phase via a senior-QE reviewer + devil's-advocate + refiner loop. Completed phases are never modified (checksum-gated). Sister skill to /draft-plan, scoped to test specs.

Orchestrate a batch bug-fixing sprint: dispatch implementers in per-issue worktrees, verify, optionally auto-land via /land-pr. Recurring via every SCHEDULE; stop/next manage it. sync updates trackers + closes already-fixed issues; plan drafts plans for skipped ones.

Helper for PR landing — rebase, push, create-or-detect PR, poll CI, optional auto-merge. Dispatched via the Skill tool by /run-plan, /commit pr, /do pr, /fix-issues pr, /draft-plan, /refine-plan, /draft-tests (and orchestrator agents landing one-off PRs). Returns state via --result-file for caller-driven fix-cycle loops. Not for direct slash invocation — humans should use /commit pr instead.

Refine an in-progress plan by reviewing remaining phases against completed work. Dispatches reviewer + devil's-advocate agents to surface stale references, invalidated assumptions, and specification gaps, then refines until convergence. Completed phases are NEVER modified. Appends a Drift Log + Plan Review.

Execute the next phase of a plan: parse status, dispatch implementation in a worktree, verify via a separate agent, update progress, write the plan report (`$ZSKILLS_REPORTS_DIR/plan-{slug}.md`), and optionally auto-land to main. Self-schedules via cron; use `next` to check, `stop` to cancel.

Batch-execute the prioritized ready queue from the dashboard: reads .zskills/monitor-state.json (plans.ready) and dispatches /run-plan auto per entry (mode resolves to `finish` by default; `phase` opts out), honoring each plan's queued mode. With `automerge`, also requests auto-merge on the PR after CI passes. N|all composes with every SCHEDULE + now for the queue-worker pattern (N plans per fire). Also manages the queue (add/rank/remove/default) and recurring schedules. Mirrors /fix-issues for bugs.

Post-PR-merge local normalization: fetch-and-prune origin, switch off merged feature branches, pull main, and delete local branches whose upstream is gone, whose PR has merged, or whose tip is fully contained in main (0 commits ahead). Preview by default — run `/cleanup-merged apply` to execute. `local` (default) / `remote` / `all` pick the scope; pass explicit branch names to narrow the candidate set. `--force` overrides the merged-check + unpushed guard for branches you explicitly name. Protected branches from config are NEVER deleted (even with `--force`) — they are always skipped.

Full pipeline: decompose a broad goal into sub-plans, draft each with adversarial review, then execute all of them autonomously via /run-plan. One command, walk away.

Decompose a broad goal into a sequence of executable sub-plans. Researches the domain, identifies sub-problems and dependencies, produces a meta-plan whose phases each delegate to /run-plan.

Install or update Z Skills supporting infrastructure (CLAUDE.md rules, hooks, scripts)

Verify all recent changes: review diffs, check that unit/e2e tests cover the changes, run all tests, manually verify UI changes with playwright-cli, fix problems, re-verify until clean, then report with recommendations.

Local web dashboard — plans, issues, worktrees, branches, tracking activity, drag-and-drop priority queue. Starts a detached Python HTTP server on a port from DEV_PORT / dev_server.default_port / port.sh; stop sends SIGTERM; restart = stop+start (for code reloads). State at .zskills/monitor-state.json.

Generate a project briefing: worktree status, open checkboxes, recent commits. Modes: summary (default), report, verify, current, worktrees. Period: 1h, 6h, 24h, 2d, 7d.

Create a git worktree for agent work. Thin wrapper around create-worktree.sh — owns prefix-derived path, optional --branch-name override, optional pre-flight prune+fetch+ff-merge, safe worktree-add, and sanitised .zskills/tracked / .zskills/worktreepurpose writes. Prints the worktree path on stdout.

Review ALL unreported sprint results: walk through manual verifications, land fixes to main, close GitHub issues, update trackers, clean up worktrees. Covers $ZSKILLS_REPORTS_DIR/SPRINT_REPORT.md and any landed-but-unclosed issues from prior sprints.

Audit what THIS session said it would do vs. what's actually shipped. Verifies session-mentioned items against ground truth (git, PRs, plans, worktrees), not conversation memory — items may have been completed in another session.

Plan dashboard. View plan status, find the next ready plan. For batch execution, see `/work-on-plans`.

QE audit: check recent commits for test coverage gaps, or bash/stress-test features to find bugs. Files GitHub issues for findings. Recurring via every SCHEDULE; stop/next manage it.

Deep debugging for one complex bug at a time. Disciplined workflow: reproduce, trace, state root cause, fix, verify. The agent must PROVE root-cause understanding before writing a fix.

Support skill that helps playwright-cli interact with a UI the way a user would — clicking menus, click-and-dragging elements on canvases, typing into inputs, using keyboard shortcuts — and documents workarounds where playwright-cli is too limited to mimic a particular user action. Dispatched when a caller (e.g. /verify-changes) needs to verify a UI change. Use when told to "test manually", "test in the browser", or "verify with playwright-cli".

A basic test fixture skill.

This skill has a description-extra block scalar whose continuation lines literally contain version: "should-not-be-redacted" text.

Fixture used to test dotfile invariance.

A fixture skill that has additional files.