做什么
使用Twill Cloud编码代理管理Twill的公用v1API工作流程. 创建/列表/更新任务,流出并取消任务,管理计划任务,列表寄存器,并导出 Claude teleport 会话.
技能库 智客分类:Agent 工作流 twill-cloud-coding-agent
使用Twill Cloud编码代理管理Twill的公用v1API工作流程. 创建/列表/更新任务,流出并取消任务,管理计划任务,列表寄存器,并导出 Claude teleport 会话.
官方网址:作者主页
先看中文介绍;官方 description 原文单独保留,不改写 SKILL.md。
使用Twill Cloud编码代理管理Twill的公用v1API工作流程. 创建/列表/更新任务,流出并取消任务,管理计划任务,列表寄存器,并导出 Claude teleport 会话.
最新计划工作完成并准备审批
按 Agent Skills 渐进披露:启动时只加载 name 与 description(约 100 token);任务匹配后才读入整份 SKILL.md 正文;scripts/、references/、assets/ 仅在需要时再读。 本文件正文结构:Twill Cloud Coding Agent、Setup、Endpoint Coverage (Public v1)、Auth and Discovery、Tasks、Create Task。
文件分析:这是一份仅含 SKILL.md 的指令型技能,代理激活后整份正文进入上下文。
Use Twill Cloud Coding Agent to manage Twill's public v1 API workflows. Create/list/update tasks, stream and cancel jobs, manage scheduled tasks, list repositories, and export Claude teleport sessions.
Twill Cloud Coding AgentSetupEndpoint Coverage (Public v1)Auth and DiscoveryTasksCreate TaskList TasksGet Task DetailsSend Follow-Up MessageList Task JobsApprove PlanCancel Task
兼容:Requires access to https://twill.ai/api/v1, curl, and a TWILL_API_KEY environment variable.
来源分类:skills.sh agent-skill
nametwill-cloud-coding-agentdescriptioncompatibility具体调用语法与可用工具以目标 Agent 客户端为准。 查看调用机制说明 ↗
先选择目标 Agent 和安装范围,保留技能包的附属文件,安装后检查客户端能否发现该技能。
复制安装指令给支持 Agent Skills 的代理,确认其中的目标目录与客户端匹配。
把 Agent Skill「twill-cloud-coding-agent」安装到我的项目:SKILL.md 原文与官方 description 见 https://zicq.com/zh/skills/skl-87b0b9e942f5b909-%E5%B0%86%E4%BA%91%E7%BC%96%E7%A0%81%E4%BB%A3%E7%90%86.html 请存为 .cursor/skills/twill-cloud-coding-agent/SKILL.md 或 .claude/skills/twill-cloud-coding-agent/SKILL.md,frontmatter 的 name 与 description 保持原样,不要改写。
需要 Node.js 与 npx。先查看仓库技能列表,确认实际名称。
npx skills add 'https://github.com/twillai/skills' --list
npx skills add 'https://github.com/twillai/skills' --skill 'twill-cloud-coding-agent'
CLI 会交互选择目标 Agent,默认安装到项目;用户级安装使用 -g。先通过查看命令核对仓库内容,再用 npx skills list 检查已安装技能。
Use this skill to run Twill workflows through the public v1 API.
Set API key and optional base URL:
export TWILL_API_KEY="your_api_key"
export TWILL_BASE_URL="${TWILL_BASE_URL:-https://twill.ai}"
All API calls use:
Authorization: Bearer $TWILL_API_KEY
Use this helper to reduce repetition:
api() {
curl -sS "$@" -H "Authorization: Bearer $TWILL_API_KEY" -H "Content-Type: application/json"
}
GET /api/v1/auth/meGET /api/v1/repositoriesPOST /api/v1/tasksGET /api/v1/tasksGET /api/v1/tasks/:taskIdOrSlugPOST /api/v1/tasks/:taskIdOrSlug/messagesGET /api/v1/tasks/:taskIdOrSlug/jobsPOST /api/v1/tasks/:taskIdOrSlug/approve-planPOST /api/v1/tasks/:taskIdOrSlug/cancelPOST /api/v1/tasks/:taskIdOrSlug/archiveGET /api/v1/tasks/:taskIdOrSlug/teleport/claudeGET /api/v1/jobs/:jobId/logs/streamPOST /api/v1/jobs/:jobId/cancelGET /api/v1/scheduled-tasksPOST /api/v1/scheduled-tasksGET /api/v1/scheduled-tasks/:scheduledTaskIdPATCH /api/v1/scheduled-tasks/:scheduledTaskIdDELETE /api/v1/scheduled-tasks/:scheduledTaskIdPOST /api/v1/scheduled-tasks/:scheduledTaskId/pausePOST /api/v1/scheduled-tasks/:scheduledTaskId/resumeValidate key and workspace context:
curl -sS "$TWILL_BASE_URL/api/v1/auth/me" -H "Authorization: Bearer $TWILL_API_KEY"
List available GitHub repositories for the workspace:
curl -sS "$TWILL_BASE_URL/api/v1/repositories" -H "Authorization: Bearer $TWILL_API_KEY"
api -X POST "$TWILL_BASE_URL/api/v1/tasks" -d '{"command":"Fix flaky tests in CI","userIntent":"SWE"}'
Repository and branch are no longer accepted in the request body — the task runs against the workspace's connected repos as resolved by the agent at runtime.
Required fields:
commandOptional fields:
agent (provider or provider/model, for example codex or codex/gpt-5.4; provider shorthands accepted: claude-code, codex, open-code)userIntent (SWE, PLAN, ASK, DEV_ENVIRONMENT) — defaults to SWEreasoningEffort (low, medium, high, xhigh)parentId (id of an existing task to spawn this one from)titlefiles (array of { filename, mediaType, url })Response is { task: { id, slug, title, url }, job: { id, status } }. Always report task.url back to the user.
curl -sS "$TWILL_BASE_URL/api/v1/tasks?limit=20&cursor=BASE64_CURSOR" -H "Authorization: Bearer $TWILL_API_KEY"
Supports cursor pagination via limit (default 20, max 100) and cursor. Response is { tasks, nextCursor }; each task includes id, slug, title, url, createdAt, latestJobStatus, and pr.
curl -sS "$TWILL_BASE_URL/api/v1/tasks/TASK_ID_OR_SLUG" -H "Authorization: Bearer $TWILL_API_KEY"
Returns task metadata plus latestJob including id, status, type, agentProvider, startedAt, completedAt, plan, planOutcome, finalAnswer (useful for ASK jobs), and pr when available.
api -X POST "$TWILL_BASE_URL/api/v1/tasks/TASK_ID_OR_SLUG/messages" -d '{"message":"Please prioritize login flow first","userIntent":"PLAN"}'
Sending a message cancels any in-flight job for the task and starts a fresh run with this message (API/CLI bypass the UI queueing flow).
Optional fields: userIntent, reasoningEffort (low, medium, high, xhigh), files.
curl -sS "$TWILL_BASE_URL/api/v1/tasks/TASK_ID_OR_SLUG/jobs?limit=30&cursor=BASE64_CURSOR" -H "Authorization: Bearer $TWILL_API_KEY"
Supports cursor pagination:
limit defaults to 30 (max 100)cursor fetches older pagesjobs and nextCursorUse when the latest plan job is completed and ready for approval.
api -X POST "$TWILL_BASE_URL/api/v1/tasks/TASK_ID_OR_SLUG/approve-plan" -d '{}'
api -X POST "$TWILL_BASE_URL/api/v1/tasks/TASK_ID_OR_SLUG/cancel" -d '{}'
api -X POST "$TWILL_BASE_URL/api/v1/tasks/TASK_ID_OR_SLUG/archive" -d '{}'
curl -N "$TWILL_BASE_URL/api/v1/jobs/JOB_ID/logs/stream" -H "Authorization: Bearer $TWILL_API_KEY" -H "Accept: text/event-stream"
Stream emits JSON payloads in data: lines and terminates with a complete event.
api -X POST "$TWILL_BASE_URL/api/v1/jobs/JOB_ID/cancel" -d '{}'
curl -sS "$TWILL_BASE_URL/api/v1/scheduled-tasks" -H "Authorization: Bearer $TWILL_API_KEY"
api -X POST "$TWILL_BASE_URL/api/v1/scheduled-tasks" -d '{
"title":"Daily triage",
"message":"Review urgent issues and open tasks",
"cronExpression":"0 9 * * 1-5",
"timezone":"America/New_York",
"agentProviderId":"claude-code/sonnet"
}'
Required: title (max 200 chars), message, cronExpression.
Optional: timezone (IANA name, defaults to "UTC"), agentProviderId (provider/model override, e.g. claude-code/sonnet, codex/gpt-5.4).
Repository / branch are no longer part of the scheduled-task payload — each run resolves repos at agent dispatch time, the same way one-shot tasks do.
Response is { scheduledTask: { id, workspaceId, createdById, title, message, cronExpression, timezone, nextRunAt, lastRunAt, enabled, agentProviderId, createdAt, updatedAt } }.
curl -sS "$TWILL_BASE_URL/api/v1/scheduled-tasks/SCHEDULED_TASK_ID" -H "Authorization: Bearer $TWILL_API_KEY"
api -X PATCH "$TWILL_BASE_URL/api/v1/scheduled-tasks/SCHEDULED_TASK_ID" -d '{
"message":"Updated instructions",
"cronExpression":"0 10 * * 1-5",
"agentProviderId":"codex/gpt-5.4"
}'
curl -sS -X DELETE "$TWILL_BASE_URL/api/v1/scheduled-tasks/SCHEDULED_TASK_ID" -H "Authorization: Bearer $TWILL_API_KEY"
PATCH accepts any subset of title, message, cronExpression, timezone, agentProviderId (pass null to clear the override). DELETE returns { "success": true }.
api -X POST "$TWILL_BASE_URL/api/v1/scheduled-tasks/SCHEDULED_TASK_ID/pause" -d '{}'
api -X POST "$TWILL_BASE_URL/api/v1/scheduled-tasks/SCHEDULED_TASK_ID/resume" -d '{}'
userIntent (SWE, PLAN, ASK, DEV_ENVIRONMENT) when calling API endpoints directly.repository / branch on tasks or repositoryUrl / baseBranch on scheduled tasks — these fields were removed; repo/branch is resolved by the agent at run time from workspace context.task.url, and only poll/stream logs when requested.TWILL_API_KEY if missing.Agent 工作流
创造有效技能指南。 当用户想创造出新的技能(或更新现有的技能),以专业知识,工作流程,或工具集成来扩展克洛德的能力时,应该使用这种技能.
Agent 工作流
使用ClawdHub CLI搜索,安装,更新并发布从taladhub.com的代理技能. 需要获取苍蝇上的新技能时使用,将安装的技能同步到最新版本或特定版本,或者发布 npm-instainddhub CLI 的新/更新的技能文件夹.
Agent 工作流
管弦乐团多代理团队,任务设定周期,交接协议,审查工作流程. 使用时间: (1)建立2+特派员队伍,具有不同专业,(2)确定任务路线和生命周期(收录框_ spec_建设_审查_完成),(3)在特派员之间制定交接协议,(4)建立审查和质量关口,(5)管理特派员之间的交流和文物共享.
Agent 工作流
Spec-first,TDD,子代理驱动的软件开发工作流程. 当:(1)构建任何新功能或应用——触发脑暴_计划_子代理执行回路,(2)调试出一个bug或测试失败——触发系统性的根起过程,(3)用户说"让我们构建","帮助我计划","我想添加X",或"这个被打破",(4)完成一个功…