菜单
Manage scheduled, background agentic tasks. Use when users ask to create, list, inspect, or delete schedules, recurring agent tasks, reminders, or background dispatch of natural-language instructions.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/jingkaihe/skills --skill schedule命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
基于 SOC 职业分类
正在显示 SKILL.md
| name | schedule |
| description | Manage scheduled, background agentic tasks. Use when users ask to create, list, inspect, or delete schedules, recurring agent tasks, reminders, or background dispatch of natural-language instructions. |
Use the skill-local scheduler command to manage background agentic work. Run commands from this skill directory through the wrapper script:
skills/schedule/scripts/agentic-schedule <command> [options]
The wrapper executes a uv project in skills/schedule, so dependencies are declared in skills/schedule/pyproject.toml and the Python source lives under skills/schedule/src/.
| Command | Purpose |
|---|---|
list | List schedules |
get NAME | Get one schedule by name |
create NAME --when WHEN --instruction INSTRUCTION | Create or replace a schedule |
delete NAME | Delete one schedule by name |
start | Install and start a persistent user-level scheduler daemon |
stop | Stop the scheduler daemon without removing its user-level config |
uninstall | Stop the scheduler daemon and remove its user-level config |
status | Show daemon, dispatcher, and schedule counts |
Examples:
skills/schedule/scripts/agentic-schedule list
skills/schedule/scripts/agentic-schedule status
skills/schedule/scripts/agentic-schedule start
skills/schedule/scripts/agentic-schedule stop
skills/schedule/scripts/agentic-schedule uninstall
skills/schedule/scripts/agentic-schedule get daily-repo-health
skills/schedule/scripts/agentic-schedule delete daily-repo-health
Use this skill when the user wants an agentic task to happen later or repeatedly, for example:
Use skills/schedule/scripts/agentic-schedule create with:
NAME: stable snake/kebab/dot-style name using only letters, numbers, ., _, -.--instruction: natural-language task for the scheduled agentic work. Put only the work to perform here; do not add any preamble about the schedule or timing.--harness: agent harness to use, one of kodelet, claude, or codex. Defaults to SCHEDULE_SKILL_HARNESS if set, otherwise kodelet.--when: one of the supported schedule expressions below.--timezone: IANA name such as America/New_York for timezone-less times.--working-directory: where the scheduled task should run. Defaults to the current working directory.--retention: how long to keep finished run records and logs. Defaults to 5d; supports compact durations such as 30d, 30h, and 30min.--overwrite to replace an existing schedule.--env KEY=VALUE for extra environment variables.The scheduler intentionally hides execution flags, so agents only need to decide the natural-language instruction and timing. Keep timing exclusively in --when; keep --instruction focused on the task itself.
Harness execution commands are fixed internally and include JSON/JSONL streaming where available:
| Harness | Command |
|---|---|
kodelet | kodelet run --headless INSTRUCTION |
claude | claude --dangerously-skip-permissions -p --output-format stream-json INSTRUCTION |
codex | codex --yolo exec --json INSTRUCTION |
Instruction examples:
Good — timing is only in --when; --instruction contains only the work:
skills/schedule/scripts/agentic-schedule create bbc-news-digest \
--when "daily 09:00" \
--instruction "Fetch the top 10 stories from bbc.com. Create a concise Markdown digest with each headline, a 1-2 sentence summary and link. Write it to ~/news/bbc-com-YYYY-MM-DD."
Bad — the instruction repeats schedule/timing context:
skills/schedule/scripts/agentic-schedule create bbc-news-digest \
--when "daily 09:00" \
--instruction "At the scheduled time, fetch the top 10 stories from bbc.com. Create a concise Markdown digest with each headline, a 1-2 sentence summary and link. Write it to ~/news/bbc-com-YYYY-MM-DD."
Supported --when formats:
| Format | Meaning |
|---|---|
now | Run once immediately |
in 90 minutes | Run once after a dateparser-readable relative time |
2026-05-25T14:00:00-04:00 | Run once at an absolute time |
every 1 hour 30 minutes | Run repeatedly at that interval; first run after the interval |
every 1d starting now | Run repeatedly, first run now |
every 2h starting 2026-05-25T15:00:00Z | Run repeatedly from an anchor time |
daily 09:00 | Run every day at local or specified timezone time |
weekly monday 09:00 | Run every week at local or specified timezone time |
One-time dates and interval durations are parsed with dateparser, so natural relative phrases such as in 90 minutes work. Daily and weekly recurring schedules remain explicit wrappers around the parsed time of day.
If the user gives a relative or informal time (“tomorrow morning”, “every weekday”), translate it to one of these explicit expressions before calling the command. If the requested cadence cannot be represented exactly, explain the nearest supported option and ask a narrow clarification.
Create example:
skills/schedule/scripts/agentic-schedule create daily-repo-health \
--instruction "Inspect this repository for failing tests or obvious regressions and summarize what needs attention." \
--when "daily 09:00" \
--timezone America/New_York
skills/schedule/scripts/agentic-schedule start to install and start a persistent user-level daemon. This uses systemd --user on Linux or LaunchAgents on macOS and does not require sudo.skills/schedule/scripts/agentic-schedule status to inspect daemon state, dispatcher state, and active schedule counts.skills/schedule/scripts/agentic-schedule stop to stop the daemon while keeping its config installed.skills/schedule/scripts/agentic-schedule uninstall to stop the daemon and remove its user-level config. This does not delete stored schedules or logs.start to run due schedules.skipped instead of executing stale work.5d.AGENTIC_SCHEDULE_DIR to use a specific state directory.AGENTIC_SCHEDULE_POLL_SECONDS to change dispatcher polling frequency; default is 5 seconds.SCHEDULE_SKILL_HARNESS to choose the default harness for newly created schedules; explicit create --harness wins.start again so the persistent daemon picks up the updated environment.logs/<schedule-name>/.For reliable reboot/login behavior, install the user-level daemon with start.
Schedule output is JSON and includes next run times, recent run status, log paths, and dispatcher status. Environment values are redacted by default. Only pass --include-environment to list or get if the user explicitly needs to inspect stored environment values.
Deleting a schedule does not stop already-running work that was launched before deletion.
Use whenever working with Gmail, Google Calendar, Google Contacts, or Google Drive through the google-workspace MCP server. Covers the maco Python wrapper workflow, tool discovery, account selection, Gmail queries, attachments, event timing, Drive exports, and safe mutations.
Run AI agents and arbitrary code in ephemeral micro-VMs with VM-level isolation, network allowlisting, and host-side secret injection. Use for Matchlock CLI and SDK workflows.
Run, benchmark, and serve local GGUF / llama.cpp models on AMD Strix Halo systems such as Framework Desktop using kyuz0 toolboxes, unified-memory sizing, and reproducible podman commands. Use this skill for Strix Halo local LLM work, especially llama.cpp setup, memory-fit or performance tuning, toolbox benchmarking, and podman-based serving.