From t-tools
Executes phased task plans by dispatching work to sub-agents for backend, frontend, miniapp, or demo phases. Reads task state files and serializes item execution from DAG.
How this skill is triggered — by the user, by Claude, or both
Slash command
/t-tools:t-run [任务名称] [--phase <backend|frontend|miniapp|demo>][任务名称] [--phase <backend|frontend|miniapp|demo>]This skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
运行时边界统一参考:`${CLAUDE_PLUGIN_ROOT}/protocols/runtime-boundaries.md`
运行时边界统一参考:${CLAUDE_PLUGIN_ROOT}/protocols/runtime-boundaries.md
执行时以当前 item 的成功标准和最小必要验证为目标;遇到冲突、缺失上下文或无法判断的语义问题时停止并说明。
上游输入(来自 /t-task 产出):
.ai/task/[feature]/.state.json — 任务状态文件(必须存在且可解析)
phases[phase].generated_at 非空).ai/task/[feature]/<phase>/index.md — 阶段总览.ai/task/[feature]/<phase>/<slot>.md — Slot manifest.ai/task/[feature]/<phase>/<slot>/<ITEM-ID>-*.md — Item 文件前置阶段状态要求:
frontend 依赖 backend == completedminiapp 是可选阶段,启用时依赖 frontend == completeddemo 依赖 active phases 中排在它之前的最后一个交付阶段 completed下游产出:
.state.json — item/slot/phase 状态变更
completed_at 和 handoff_summarylast_error.ai/task/[feature]/.state.json。/t-run 的执行单元、slot 顺序、失败处理、所需上下文统一参考 ${CLAUDE_PLUGIN_ROOT}/protocols/task-phase-execution.md。index.md 和 slot manifest 只作为上下文和导航,不作为直接执行输入。finalize.md 不由 /t-run 执行。accept slot 完成后,/t-run 必须停止并提示执行 /t-backend-finalize [feature];不得自动执行 finalize.md。| 参数 | 说明 |
|---|---|
[feature] | 功能名 |
--phase <backend|frontend|miniapp|demo> | 仅执行指定阶段;未指定时执行 .state.json 的当前阶段 |
.ai/task/[feature]/.state.json 必须存在且可解析。--phase miniapp。phases[phase].generated_at 非空。${CLAUDE_PLUGIN_ROOT}/protocols/task-phase-execution.md。index.mddev.md, test.md, accept.md;demo 为 dev.md, accept.mdfinalize.md${CLAUDE_PLUGIN_ROOT}/protocols/task-state-contract.md${CLAUDE_PLUGIN_ROOT}/protocols/task-phase-execution.md按 ${CLAUDE_PLUGIN_ROOT}/protocols/task-phase-execution.md 选择可执行 item:
pending 或 failed item/t-task-check每个 item 必须通过 Agent tool 启动,subagent_type 为 item 文件中的 agent 字段值。传入 prompt 必须包含最小上下文(见下方),agent 规范文件路径作为指令引用。
最小上下文、可选增强上下文以及 backend-test 额外要求统一参考:
${CLAUDE_PLUGIN_ROOT}/protocols/task-phase-execution.md${CLAUDE_PLUGIN_ROOT}/protocols/backend-test-execution.mdbackend/test 特例:
test_item_type,只允许 authoring 或 runner。test_item_type 时拒绝执行,提示先运行 /t-task-check 或重建/修正 item。authoring:不加载 t-backend-test-run,只编写或调整场景测试并做编译验证。runner:加载 ${CLAUDE_PLUGIN_ROOT}/skills/t-backend-test-run/SKILL.md,在全部相关 authoring item 完成后集中执行定向测试、失败分类、生产代码修复委派和重测。failed,下次 /t-run 重试failed 并注明 compilation cascadehandoff_summary,可降级启动,但必须显式标注 handoff 缺失${CLAUDE_PLUGIN_ROOT}/protocols/task-state-contract.md 与 ${CLAUDE_PLUGIN_ROOT}/protocols/task-phase-execution.md 校验状态与 DAG。running:
.state.json 后再重试。tasks[phase][slot].items[item_id].status = runningtasks[phase][slot].items[item_id].started_at = <timestamp>tasks[phase][slot].status = runningphases[phase].status = runningtasks[phase][slot].items[item_id].status = completedtasks[phase][slot].items[item_id].completed_at = <timestamp>tasks[phase][slot].items[item_id].handoff_summary = <summary>tasks[phase][slot].items[item_id].status = failedtasks[phase][slot].items[item_id].last_error = <summary>tasks[phase][slot].status = failedphases[phase].status = failedaccept slot 全部 completed 后停止,并提示执行 /t-backend-finalize [feature]。dev.md、test.md、accept.md。index.md 或 slot manifest 就开始执行。backend-test 直接下发"先跑全量 uv run scripts/backend-test.py --"而不做变更分析。accept 完成后,不得自动执行 finalize.md。/t-task [feature] --phase [phase]。.state.json.phases 中的 active phases。/t-task [feature] --phase [phase]。# 按阶段执行
/t-run sample-feature --phase backend
# 调用 backend-dev item 时的最小上下文
feature: sample-feature
phase: backend
slot: dev
item_id: BE-D02
agent: backend-dev
item_file: .ai/task/sample-feature/backend/dev/BE-D02-domain-models.md
slot_manifest: .ai/task/sample-feature/backend/dev.md
phase_index: .ai/task/sample-feature/backend/index.md
dependencies:
BE-D01: completed, handoff=<summary>
npx claudepluginhub timzaak/web-dev-skillsExecutes implementation plans by dispatching tasks to implementer and reviewer subagents. Tracks progress per-task and coordinates sequential phases.
Converts technical design documents into executable phased task plans with work breakdown, dependencies, and state tracking.
Executes implementation plans with automatic sequential/parallel orchestration, handling worktree verification, resume detection, phase dispatch, and quality verification.