// Universal task dispatcher. Start, route, and execute any task through the development workflow (Steps 0-9). Invoke on every task — /task <description>, /task
[HINT] Download the complete skill directory including SKILL.md and all related files
| name | task |
| description | Universal task dispatcher. Start, route, and execute any task through the development workflow (Steps 0-9). Invoke on every task — /task <description>, /task |
| version | 1.0.0 |
Before starting, pick the right tool for the job:
| Activity | Tool | Why |
|---|---|---|
| Architecture planning | Claude Code /plan | Opus, 1M context |
| Feature implementation | Claude Code | Hooks, subagents, worktrees |
| Quick single-file fix | VS Code Claude ext | Faster for small scope |
| DevOps / terminal | Codex CLI | Strong at terminal tasks |
| Background refactor | Codex Cloud | Fire-and-forget PR |
| Code review | GitHub Actions | Async, no session cost |
| Test writing | subagent: test-writer (CC) / inline (Codex) | Isolated context |
| Documentation | subagent: doc-writer (CC) / inline (Codex) | Cheap, focused |
| Research | Claude.ai Projects | Persistent, web search |
This skill uses Claude Code features. Portable equivalents for Codex/Cursor:
| Claude Code | Codex / Cursor |
|---|---|
TaskCreate / TaskUpdate | Markdown checklist in .branch-context.md |
EnterWorktree | Use absolute paths + git -C for all ops |
/orient, /brainstorming, etc. | Read .agents/skills/<name>/SKILL.md and follow manually |
Subagent dispatch (implementer, test-writer) | Execute sequentially in main session |
| Agent Teams | Not supported — use sequential work |
/review, /plan-review | Read .agents/skills/review/SKILL.md and follow manually |
| Invocation | Behavior |
|---|---|
/task <description> | Start new task |
/task #<issue> | Start from GitHub issue |
/task | Start new task — ask what to work on |
In copilot mode (operating-mode.md), the workflow is a menu, not a pipeline:
operating-mode.md explicitly overrides.These rules are agent-interpreted guidance, not machine-enforced. They cannot be skipped regardless of task type or scope unless noted. Omission is not permission — any step not explicitly listed as skippable is required.
REQUIRE task_list.all_steps_populated BEFORE step_1
REQUIRE worktree BEFORE code
REQUIRE orient.complete BEFORE plan
REQUIRE tests BEFORE implementation # TDD: RED → GREEN → REFACTOR
REQUIRE review(plan).pass BEFORE build UNLESS scope == one_sentence OR task_type IN [fix, hotfix]
REQUIRE validate.pass BEFORE review(build)
REQUIRE review(build).pass BEFORE archive
REQUIRE task_list.no_pending_steps BEFORE ship
REQUIRE branch.base IS CURRENT(origin/main) UNLESS stacking
REQUIRE plan.commit BEFORE review(plan)
REQUIRE batch.commit BEFORE batch.next
WHEN ci.down:
USE local_merge AT step_8
WHEN files_touched MATCH [AGENTS.md, config.toml, sync.sh, skills]:
REQUIRE codex_specialist IN review_panel
Persist significant decisions to .branch-context.md during Steps 3-5 — not just at reflect/bail time.
Create a task list with ALL steps for the detected path. Every step is a task — no exceptions. The task list is the enforcement mechanism: pending tasks are visible proof of unfinished required work.
TaskCreate one task per step. Populate all steps upfront (adjust after orient detects path). Mark each in_progress before starting, completed when done..branch-context.md with all steps.Invoke /orient with args (Codex: read .agents/skills/orient/SKILL.md). It produces:
After orient completes, populate the task list with remaining steps for the detected path:
fix/*, hotfix/*): ALWAYS skip Steps 3-4 regardless of scope. No plan, no plan review — go directly to Step 5.Invoke /isolate with the branch name from Step 1 (Codex: read .agents/skills/isolate/SKILL.md). It handles freshness (fetch origin main), worktree creation from origin/main, directory selection, gitignore safety, dependency install, and baseline test verification. If already in a worktree, it freshens the branch instead of creating a new one.
Path discipline: cd does not persist between Claude Code Bash calls. After worktree creation, use absolute paths for all file operations and git -C <absolute-path> for git commands. Never rely on cd carrying over.
Hotfix (fix/*, hotfix/*): SKIP — proceed to Step 5. No architect agent, no plan file. Add a one-paragraph diagnosis to the commit message instead. Log: "Step 3 — Plan: SKIPPED (hotfix fast path)."
One-sentence scope: SKIP — proceed to Step 5. Log: "Step 3 — Plan: SKIPPED (one-sentence scope)."
Full scope:
3a. Brainstorm. Invoke /brainstorming with the task description. One round — refine the idea, confirm constraints, explore 2-3 approaches.
3b. Write plan. Write a plan to ai-workspace/plans/<name>.md using the template at ai-workspace/plans/TEMPLATE.md. Fill all template fields including:
Do NOT review the plan here — that's Step 4.
No plan (hotfix or one-sentence scope): SKIP — proceed to Step 5.
Has plan:
Run /review against the plan file (Codex: read .agents/skills/review/SKILL.md and follow manually).
technical-editorarchitect-reviewercodex-specialistsecurity-auditorui-designerOn APPROVE: proceed to Step 5. On ESCALATE after 3 rounds: stop and present unresolved findings to human.
Invoke /build (Codex: read .agents/skills/build/SKILL.md). It handles plan parsing, batch-sequential/task-parallel subagent dispatch, TDD, inter-batch regression checks, and the validate hard gate. For hotfixes or one-sentence scope, it implements directly with TDD. Step 6 cannot begin until /build reports validate passing.
Hotfix protocol (applies when task_type is fix or hotfix):
Run /review against the changed files (git diff main...HEAD) (Codex: read .agents/skills/review/SKILL.md and follow manually).
code-reviewerarchitect-reviewercodex-specialistsecurity-auditordesign-reviewerOn APPROVE: proceed to Step 7. On ESCALATE: stop and present unresolved findings to human.
Has plan:
Invoke /archive (Codex: fill Outcomes & Learnings in plan file, rename to .done.md). It fills Outcomes & Learnings in the plan file and renames to .done.md.
No plan (hotfix / one-sentence):
Ensure root cause or change rationale is documented in the commit message body. No plan file to archive.
Invoke /ship (Codex: read .agents/skills/ship/SKILL.md). It handles validation, option presentation (PR + auto-merge vs local merge vs keep vs discard), CI up/down routing, shallow-clone merge for the local path, and worktree cleanup. CI status comes from conversation context — if the user said CI is down, /ship uses the local merge path.
When files touched include AGENTS.md, config.toml, sync.sh, or skills → codex-specialist must be in the review panel (enforced at Step 6).
Invoke /reflect (Codex: read .agents/skills/reflect/SKILL.md). It handles:
.branch-context.md for session learningsMEMORY.md on mainpnpm install)pnpm test:all)pnpm typecheck)pnpm lint)pnpm format:check)/review panel)/reflect and pushed to primaryInvoke /bail at any step to abort cleanly. It handles: learnings capture to .branch-context.md, issue update, PR close if open, worktree cleanup. Bail replaces Step 9 (Reflect) — it captures learnings itself. Clean up any open task list items.