Scheduled jobs in chat
Daily recurring tasks the user wants the assistant to run automatically at a fixed clock time and deliver results to this Telegram chat. Use when the user asks for a digest, reminder, or agent task every day at a specific hour, or mentions cron-style daily schedules in natural language.
- License 许可
- MIT
- 最近更新
- 2026-09-17
- Stars
- ★ 1
- Forks
- 0
01 前言 Front Matter
- name
- Scheduled jobs in chat
- tools
- [ "create_scheduled_job" ]
- description
- Daily recurring tasks the user wants the assistant to run automatically at a fixed clock time and deliver results to this Telegram chat. Use when the user asks for a digest, reminder, or agent task every day at a specific hour, or mentions cron-style daily schedules in natural language.
03 SKILL.md 正文
When to use create_scheduled_job
- The user wants a daily job (same wall-clock time every calendar day) executed by the assistant and delivered here.
- Extract instruction (what the job should do each run) and local hour (0–23) and minute (0–59) from the user message. Pass them as separate tool arguments; do not rely on hidden server-side parsing of a single prose blob.
- If the user gives a timezone name, pass it in
timezone; otherwise omit it so the server default applies.
Behaviour
- Call
create_scheduled_jobonce per requested job with explicitinstruction,hour, andminute. - Prefer the built-in scheduler over suggesting external cron, systemd timers, or third-party job services.
- If time or task details are missing, ask one short clarifying question instead of guessing.
After creation
- Summarize the tool output (job id, schedule, next run) for the user.
- Remind that they can manage jobs with
/jobs listand related commands.
04 同仓库更多 Skills
Demo Skill
demo skill for runtime skills integration load test
Skill One
test skill for config load
Web and GitHub source research
When the user asks to investigate, summarize, or explore a public website or an open-source project on GitHub. Guides bounded use of web_fetch, web_search, and run_on_node so responses stay within useful LLM context.
Web and GitHub output hygiene
When the user asks about the web, GitHub, URLs, fetching pages, searching online, digests from news sites, or any workflow that may use web_search or web_fetch. Prefer small tool outputs over large HTML.
Memory retrieval
When the user asks about memory, tools, or skills, choose the matching retrieval tool: read_memory (date/range), search_vector_memory (memory semantics), search_vector_tool (tool knowledge), search_vector_skill (skill knowledge).