t-super-run
Plan and execute one specified delivery phase after technical design, with persisted recovery and independent read-only acceptance.
Skill by timzaak
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/timzaak-t-super-run
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: design, productivity
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md and code. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/t-super-run
What's inside
- Skill name
t-super-run- Description
- Plan and execute one specified delivery phase after technical design, with persisted recovery and independent read-only acceptance. Use to combine planning and execution; use t-task and t-run for separate plan review or item-based execution.
- License
- Apache-2.0
- Instructions
- 99 lines in SKILL.md · view on GitHub
Agent Skills spec issues
- allowed-tools: must be a space-separated string
Files (1)
- SKILL.md
SKILL.md (first lines)
# Super Run
把任务规划与阶段执行合并为一个可恢复闭环。只做 `phase -> task` 目标级规划,由当前主会话直接完成实现、测试和修复;验收按共享协议派发对应只读 accept subagent,除此之外不调用 subagent。每次调用只执行 `--phase` 显式指定的一个 phase;该 phase 完成后停止并报告剩余未完成 phase,不自动进入下一个 phase。
运行时边界统一参考:`${CLAUDE_PLUGIN_ROOT}/protocols/runtime-boundaries.md`
设计生成状态统一参考:`${CLAUDE_PLUGIN_ROOT}/protocols/design-state-contract.md`
需求与决策边界统一参考:
- `${CLAUDE_PLUGIN_ROOT}/protocols/requirement-source-contract.md`
- `${CLAUDE_PLUGIN_ROOT}/protocols/decision-continuity-contract.md`
解析 phase、选择设计输入或规划角色与执行顺序时,读取 `${CLAUDE_PLUGIN_ROOT}/protocols/task-phase-execution.md` 的 Phases、Design Inputs、Role Mapping 和 Slot Order。
规划或恢复时读取计划、状态、资源加载和验收回退规则:
- `${CLAUDE_PLUGIN_ROOT}/protocols/super-run-state-contract.md`
规划测试与验收时读取 `${CLAUDE_PLUGIN_ROOT}/protocols/verification-evidence-contract.md`,将受影响行为的验证入口、关键断言和承接位置写入现有设计测试章节/阶段计划;交付收尾按同一协议核对证据和待验证项。
## 参数
| 参数 | 说明 |
| --- | --- |
| `[feature]` | 必填;允许中文、英文、数字、空格、下划线和连字符 |
| `--phase <phase>` | 必填;本次调用只执行该 phase,完成后停止 |
## 前置条件
- `--phase` 必填;缺失时终止并提示 `--phase <phase>`,非法或不适用时按 task-phase-execution 的 Phases 规则提示可选阶段。
- feature 必须是单个目录名,拒绝路径分隔符、`.` 和 `..`。
- `.ai/design/[feature].md` 必须存在。
- 按状态协议的 Design Source Gate 校验来源;首次规划运行 `python "${CLAUDE_PLUGIN_ROOT}/scripts/check-design.py" ".ai/design/[feature].md" --require-complete --json`,恢复时核对生成状态与设计文档集合。失败时停止,不进入执行;`t-prd-check`、`t-design-check` 和 `t-task-check` 均不是强制前置。
- 不读取或修改 `.ai/task/[feature]/` 作为 super-run 状态。
- 已有状态且请求的 phase 为 `completed | skipped` 时,按 `${CLAUDE_PLUGIN_ROOT}/protocols/verification-evidence-contract.md` 核对证据适用性;仍有效则报告阶段结果和待验证项,相关输入变化时按状态协议只重新打开受影响 task 及验收,不选择其他 phase。
## 来源加载
在规划或恢复前:
1. 读取来源门禁取得的全部 `design_documents` 和现有 `.ai/super-run/[feature]/`。
2. 按 task-phase-execution 的 Design Inputs 确认当前 phase 的设计输入。
3. 读取相关 `.ai/prd/**/*.md`、`docs/prd/**/*.md`、`.ai/user-stories/**/*.md` 与 `docs/user-stories/**/*.md`,保留 draft/published 来源边界。
4. 在任何提问前读取 `.ai/decision-log/[feature].md`;存在时按需读取 `.ai/decision/[feature].md` 与 `.ai/tech-research/[feature].md`。
5. 按 task-phase-execution 的 Phases 启用规则,从设计覆盖矩阵、Operation ID、文件影响表、Decision Trace、代码和配置确定 active phases、task 闭环与真实验证入口。
6. 按 Decision Exposure Gate 分类缺口;`needs_user_answer` 未解决时不得进入实现。
不要无差别加载所有 PRD、用户故事或 guide。先通过 feature 名、设计引用和内容检索定位相关文件,再读取全文。
## 计划与状态
- 按 `${CLAUDE_PLUGIN_ROOT}/protocols/super-run-state-contract.md` 创建或更新:
- `.ai/super-run/[feature]/.state.json`
- `.ai/super-run/[feature]/[phase].md`
- 按 task-phase-execution 的 Slot Order 和 Role Mapping 规划当前 phase 的 task 顺序与角色。
- 每个 task 只规划一个责任闭环,不生成 item。
- 把校验结果的 `design_documents` 写入 super-run state;恢复规则见共享协议。
- 按 super-run-state-contract 的 Phase Plan Contract 与 Phase Resources,将当前 task 的 agent 规范、设计输入和专项资源具体路径写入计划;仅在命中对应条件时加载专项资源。
规划写清目标、成功标准、权限边界、当前角色、证据入口和停止条件,具体实现路径根据仓库事实决定。不要把 `t-task` 的细粒度 item 换一种格式复制进来,也不要用长篇过程指令占用主会话上下文。
## 执行
循环执行直到 phase 完成或进入真正阻塞:
1. 按 Design Source Gate 核对设计生成状态与文档集合;变化时先按共享协议重规划。
2. 按状态协议选择首个未完成 task;`blocked` 先核对解除条件,未解除则停止,不能跳到后续 task。
3. dev/test task:读取该 task 的 agent 规范全文及计划列出的关联文档,把它们作为主会话当前角色边界。accept task:按共享协议派发对应只读 accept subagent,并把报告结论映射回状态。
4. 写入 `in_progress`,执行交付、最小可靠验证和必要修复。
5. 写入 `completed` 与证据,重新聚合 phase,继续下一个 task。
6. 失败时先写 `failed` 与证据;能够基于新证据修复时继续闭环,否则写 `blocked` 并停止本次执行。
7. accept 拒绝时按共享协议重新打开 dev 或已规划的 test,再次验证和验收。
每次显著步骤后把完成内容、验证证据、剩余工作和 handoff 写入状态或 phase 计划,确保上下文压缩后能从文件恢复。
持续执行由本 skill 的循环推进;仍有可执行 task 时继续,不在 task 之间例行询问是否继续。中断后用同一命令从已落盘状态恢复。
## 禁止事项
- 为 dev/test 调用 `Agent`、并行 subagent 或任何以 subagent 做上下文隔离的调度;accept 只允许串行派发计划中对应的只读 accept agent,不得派发其他角色。
- 自动进入、规划或恢复未被本次调用显式请求的 phase。
- 生成 `.ai/task/` 的 manifest/item,或修改 `t-task/t-run` 状态。
- 把 accept 角色改为实现角色,或让 accept 直接修复代码。
- 跳过测试、弱化断言、忽略失败或把 `blocked` 当作阶段完成。
- 在无真实兼容约束时保留被设计明确替换的旧路径。
## 完成条件
只有同时满足以下条件才报告当前 phase 完成:
- 当前 phase 的全部 task 为 `completed | skipped`。
- 当前设计来源门禁通过且 `sources.design.documents` 与实际设计文档集合一致。
- 计划中的验证已经执行并有证据。
- accept task 给出允许进入下游的结论;整个 phase 为 `skipped` 时必须有明确不适用证据。
- `.state.json` 已按最终结果聚合。
输出当前 phase、task 状态、主要变更、验证证据和剩余未完成 phase,然后停止本次调用;剩余 phase 由用户再次显式传入 `--phase` 启动。仅当全部 active phases 已完成且按 `${CLAUDE_PLUGIN_ROOT}/protocols/verification-evidence-contract.md` 核对全部必要场景均有有效证据时,报告 feature 的 super-run 已完成;否则列出待验证项并按共享协议重新打开承接验收,不宣称交付完成。From timzaak/web-dev-skills, Apache-2.0.
More from timzaak/web-dev-skills
- 0.675score
t-doc
timzaak
Generate human-readable project documentation — tutorials, getting-started guides, architecture overviews, API references, deployment or onboarding docs — by…
SkillAFreshness: A★ 97Claude Code +9 - 0.675score
t-icon-prompt
timzaak
Turn a vague icon idea into a ready-to-paste generation prompt card through a short guided Q&A.
SkillAFreshness: A★ 97Claude Code +9 - 0.671score
t-init
timzaak
Initialize a new Rust + React full-stack project; use --extension-demo to add Playwright integration-test infrastructure to an existing WXT extension; use…
SkillAFreshness: A★ 97Claude Code +9 - 0.670score
t-review
timzaak
Review changed code for correctness bugs with parallel read-only finders and an independent 3-state verifier for each candidate; add specialist angles and a…
SkillAFreshness: A★ 97Claude Code +9 - 0.670score
t-dream
timzaak
Organize and realign PRDs, user stories, design/task docs, implementation facts, and project structure so the target project keeps a clean current context…
SkillAFreshness: A★ 97Claude Code +9 - 0.670score
t-skill
timzaak
Create or update a skill — by default inside the t-tools plugin repository; with --project inside the current target project's skill directory.
SkillAFreshness: A★ 97Claude Code +9 - 0.670score
t-how
timzaak
Interactive guide that teaches how to use the t-tools plugin.
SkillAFreshness: A★ 97Claude Code +9 - 0.669score
t-decision
timzaak
Decide whether a feature should enter the t-tools workflow before tech research or PRD.
SkillAFreshness: A★ 97Claude Code +9 - 0.667score
t-figma-assets
timzaak
Prepare Figma bitmaps, SVGs, and videos for human-guided UI implementation; identify assets, validate their types, collect bitmap evidence, convert or…
SkillAFreshness: A★ 97Claude Code +9 - 0.667score
t-design-check
timzaak
Evaluate technical design documents for implementability, completeness, and consistency with a quantitative 100-point score.
SkillAFreshness: A★ 97Claude Code +9 - 0.667score
t-figma-impl
timzaak
Implement a Figma UI in an existing frontend file, block by block; leave pending assets empty, bind prepared assets, and hand over for manual acceptance.
SkillAFreshness: A★ 97Claude Code +9 - 0.667score
t-prd
timzaak
Draft or update PRD and user stories, clarifying unresolved product choices.
SkillAFreshness: A★ 97Claude Code +9
Signals
- Stars
- 97
- Forks
- 9
- Open issues
- 0
- Last commit (this folder)
- Oct 14, 2026 (14 days ago)
- Last release
- Oct 20, 2026 (8 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- Apache-2.0
- Repository
- github.com/timzaak/web-dev-skills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from timzaak
- 0.675score
t-doc
timzaak
Generate human-readable project documentation — tutorials, getting-started guides, architecture overviews, API references, deployment or onboarding docs — by…
SkillAFreshness: A★ 97Claude Code +9 - 0.675score
t-icon-prompt
timzaak
Turn a vague icon idea into a ready-to-paste generation prompt card through a short guided Q&A.
SkillAFreshness: A★ 97Claude Code +9 - 0.671score
t-init
timzaak
Initialize a new Rust + React full-stack project; use --extension-demo to add Playwright integration-test infrastructure to an existing WXT extension; use…
SkillAFreshness: A★ 97Claude Code +9 - 0.670score
t-review
timzaak
Review changed code for correctness bugs with parallel read-only finders and an independent 3-state verifier for each candidate; add specialist angles and a…
SkillAFreshness: A★ 97Claude Code +9 - 0.670score
t-skill
timzaak
Create or update a skill — by default inside the t-tools plugin repository; with --project inside the current target project's skill directory.
SkillAFreshness: A★ 97Claude Code +9 - 0.670score
t-dream
timzaak
Organize and realign PRDs, user stories, design/task docs, implementation facts, and project structure so the target project keeps a clean current context…
SkillAFreshness: A★ 97Claude Code +9
Cite this page
t-super-run skill. Agent Mart. https://agentmart.dev/skills/timzaak-t-super-run (last updated 2026-10-24)