소스 정보
- 저장소
- Dev-Toolbelt/dev-team-agents
- 최근 소스 활동
- 2026년 8월 2일 23:31
- 감지된 SKILL.md 언어
- 영어
- 스타
- 4
- 포크
- 0
설치 방법
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
소스 파일 검토
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
메뉴
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
직접 명령은 검토 Prompt를 거치지 않습니다. 실행하기 전에 소스를 확인하세요.
npx skills add https://github.com/Dev-Toolbelt/dev-team-agents --skill backlog-template명령은 한 줄로 유지됩니다. 복사하기 전에 가로로 스크롤해 전체 내용을 확인하세요.
로컬 사본을 원하시나요? SkillsMP에서 현재 제공할 수 있는 파일을 다운로드하세요.
SOC 직업 분류 기준
SKILL.md 표시 중
| name | backlog-template |
| description | Backlog — epics, sprints, tasks, DoD, dependencies, estimates. |
Generate under docs/backlog/:
overview.md — business requirements document (the planning deliverable)epics.md — epicsdod.md — Definition of Donesprints/ — one file per sprint plus a status index (see below)Sprint files live in the sprints/ subfolder, never at the backlog root:
docs/backlog/
├── overview.md
├── epics.md
├── dod.md
└── sprints/
├── sprints.md ← index: summary + status of every sprint
├── sprint-1.md
├── sprint-2.md
└── ...
overview.md# Project: [Name]
## Objective
[One paragraph: what problem this solves, for whom, and the expected outcome]
## Stakeholders
| Role | Name / Team | Responsibility |
|------|------------|----------------|
| Product Owner | ... | Prioritization, acceptance |
| Tech Lead | ... | Architecture, code quality |
| Dev Team | ... | Implementation |
## Constraints
- **Deadline**: [date or "not defined"]
- **Budget**: [if relevant]
- **Tech restrictions**: [e.g., "must integrate with legacy X"]
- **Compliance**: [LGPD, GDPR, PCI, etc.]
## Out of Scope
- [Explicit list of what will NOT be built in this phase]
dod.md — Definition of Done# Definition of Done
A task is DONE when ALL of the following are true:
## Code
- [ ] Implements the acceptance criteria completely
- [ ] No known bugs introduced
- [ ] Code reviewed and approved
- [ ] Linters pass (no errors)
- [ ] No hardcoded secrets or sensitive data
## Tests (if project requires)
- [ ] Unit/integration tests written for new logic
- [ ] All existing tests pass
- [ ] Coverage does not decrease
## Documentation
- [ ] Public APIs documented (if applicable)
- [ ] README updated if setup steps changed
- [ ] ADR written if an architectural decision was made
## Deployment
- [ ] Works in the target environment
- [ ] No regressions in adjacent features
epics.md# Epics
## Epic 1 — [Name]
**Goal**: [What this epic delivers]
**Acceptance Criteria**:
- [ ] ...
- [ ] ...
**Estimated effort**: [S / M / L / XL]
**Priority**: [High / Medium / Low]
**Dependencies**: [Epic X, external system Y]
---
## Epic 2 — [Name]
...
sprints/sprint-<n>.mdOne file per sprint, in the sprints/ subfolder: sprints/sprint-1.md, sprints/sprint-2.md, etc. Never merge multiple sprints into one file.
# Sprint <n> — [Theme]
**Goal**: [What should be achievable at the end of this sprint]
**Start**: YYYY-MM-DD
**End**: YYYY-MM-DD
**Estimated duration**: N weeks
**Agents**: [list of agents that will act in this sprint — see Agent Assignment table below]
## Parallel Execution Plan
Tasks are grouped into **waves**. Every task in the same wave is mutually
independent and can be built **in parallel**, each in its own git worktree
(and its own isolated Docker stack when the project uses Docker — see below).
A wave starts only after every task in the previous wave is done.
| Wave | Tasks (parallel) | Blocks (why the next wave waits) |
|------|------------------|----------------------------------|
| A | TASK-001, TASK-002 | Wave B needs the schema from TASK-001 |
| B | TASK-003, TASK-004 | — |
**Isolation model:**
- **Worktree per task** — always. Each task gets its own branch/worktree so
parallel tasks never touch the same working tree.
- **Isolated infra per worktree** — **only when the project uses Docker**
(a compose file exists). Then each worktree runs its own namespaced Docker
Compose stack (isolated containers, volumes, networks, host ports), so
parallel tasks don't collide on services or ports. **No Docker → skip this;
parallelism is achieved by worktree alone.**
## Tasks
### TASK-001 — [Name]
**Epic**: Epic 1
**Type**: [Feature | Fix | Chore | Spike]
**Agent**: [agent name — see Agent Assignment table]
**Estimate**: [hours or story points]
**Wave**: A
**Worktree branch**: [suggested `<context>/<brief-title>`, e.g. `feat/user-schema`]
**Spec**: [`docs/specs/<feature>.md` (+ `-contract.md` if the gate fired), or "none — see spec-gate skip reason"]
: [TASK-XXX or "none"]
: [ ] Pending
:
[What needs to be done — technical enough for a developer to start]
:
[ ] ...
[ ] ...
---
...
| Tasks | Count |
|-------|-------|
| Total | N |
| Feature | N |
| Fix | N |
| Chore | N |
| Waves | N |
| Max parallelism | N tasks (largest wave) |
| Estimated total | Nh / N pts |
sprints/sprints.md — status indexA single index at docs/backlog/sprints/sprints.md summarizing every sprint and its status. Create it when the first sprint is generated and keep it current — flip a sprint's status as it moves forward, and set it to Done in the same pass that finalizes the sprint.
# Sprints
| Sprint | Theme | Goal (one line) | Status |
|--------|-------|-----------------|--------|
| [sprint-1](sprint-1.md) | [Theme] | [What it delivers] | Planned |
| [sprint-2](sprint-2.md) | [Theme] | [What it delivers] | Planned |
Status values: Planned → In progress → Done.
When generating sprint files, use this table to assign agents to tasks. Multiple agents can be listed when tasks require collaboration.
| Task type | Primary agent | Secondary agent |
|---|---|---|
| API endpoint / service / business rule | backend-developer | database-specialist (if schema involved) |
| Frontend screen / component / UI flow | frontend-developer | ui-ux-designer (if new UI pattern) |
| Database schema / migration / query optimization | database-specialist | backend-developer |
| Design system / UX flow / visual spec | ui-ux-designer | — |
| CI/CD / Docker / infra / deploy | devops-specialist | — |
| Backend tests | backend-test-specialist | — |
| Frontend tests | frontend-test-specialist | — |
| Security audit / vulnerability fix | security-specialist | software-architect |
| Architecture decision / refactor | software-architect | relevant developer |
| Documentation / changelog / runbook | technical-writer | — |
| Full-stack feature | backend-developer + frontend-developer | database-specialist (if schema) · ui-ux-designer (if UI) |
At the top of each sprint file, list all agents that will be active in that sprint under the **Agents**: field.
Use these benchmarks as starting points — adjust based on team familiarity with the stack:
| Size | Description | Typical range |
|---|---|---|
| XS | Config change, minor text fix | 0.5–1h |
| S | Simple CRUD endpoint or UI form | 2–4h |
| M | Feature with business logic + tests | 4–8h |
| L | Complex feature, multiple components | 1–3 days |
| XL | Epic-level effort, architectural impact | 1–2 weeks |
Always add buffer: multiply raw estimate × 1.3 for unknowns.
When generating sprint estimates, also provide:
In task descriptions, always call out upstream dependencies:
Depends on: TASK-003 — cannot start until TASK-003 is doneDepends on: external API contract — blocked by third partyParallel with: TASK-005 — can run concurrentlyDesign for parallelism. Keep the dependency graph as flat as possible so
each wave holds as many tasks as it can. A task with Depends on: none belongs
to Wave A; a task moves to a later wave only when it truly cannot start before
another task finishes. When two tasks would collide (same file, same schema),
prefer splitting the work so they land in the same wave over serializing them —
only serialize when the dependency is real.