Skill, then routine

Skill packs › Skill

Skill, then routineIN EVERY PACK

Save a finished method as a skill before scheduling it. Use when the user asks to automate, schedule, create a routine, Cursor Automation, or Cowork scheduled task; after a successful one-off job; or when Grok Bot, Claude, and Cursor names for the same job rung would otherwise drift.

Skill file skill-then-routine.md · last updated Aug 18, 2026

How to run it. Download any pack from the skill-pack directory, unzip it into your Claude project folder, and this skill is one of the files inside. You do not paste it anywhere — the agent reads it when the job calls for it. This one ships in every pack.

Use this when a task worked once and someone wants it on a clock, or when Claude / Grok Bot / Cursor names would fork the SOP. Public map: https://blitzmetrics.com/skills-and-routines/

A skill is how. A routine (Claude: scheduled task. Cursor: Automation. Grok Bot: routine) is when. Do not schedule a demo.

Official Grok Bot: https://docs.x.ai/grok-bot/skills-routines-and-automations Job rung: https://blitzmetrics.com/persistent-agents/ Claude paste examples: https://blitzmetrics.com/18-scheduled-tasks-every-agency-owner-should-build-in-claude/

Inputs

  • One completed run (or a written method that has already passed)
  • Owning role / Bot / conversation
  • Schedule or event matcher, plus timezone
  • Approval boundary (send / spend / delete / publish / production)
  • Missing-data policy

Steps

  1. Confirm the job passed once on current inputs. If it did not, do it once. Stop.
  2. Write or update the skill in a file we own (DealCon-Skills, GitHub blitzmetrics-skills, Cursor SKILL.md). Six parts: when to use it, inputs and access, sequence, how to validate, what to return, what needs approval.
  3. Do not let a Bot bio, a chat, or a Custom Instructions box become the only copy.
  4. On Grok Bot: Settings → Plugins; enable the skill for this Bot if it is missing from /. Teach-a-task (when available) is a skill draft — add decision rules before you trust it.
  5. Test run with safe inputs. A test run is real work. Confirm it stopped at the approval gate and failed loudly on missing data.
  6. Only then create the schedule:
  • Claude: “Create a scheduled task: … run my {skill} … Do not send. If the source is missing, report failure.”
  • Grok Bot: ask the owning Bot. View conversation details → Routines. Cap 50. 20 run records. Laptop may be closed.
  • Cursor: Cursor Automation only after the skill exists. Do not open the Automations editor as a substitute for the skill file.
  1. Event triggers (Slack, GitHub) are account integrations, not the plugin. Narrow matchers. Never “every new message.”
  2. Re-test after a website, connector, or source format changes.

Definition of done

  • [ ] Skill file exists outside the chat
  • [ ] Test run (or equivalent) passed on safe inputs
  • [ ] Schedule/routine names owner, timezone, source, result, approval, missing-data policy
  • [ ] Send / spend / delete / publish stay behind a yes
  • [ ] Public map linked if you are teaching the rule

Do not

  • Enable a pile of new Grok Bot routines while that weekly meter is already empty (this account: Extra Credits after buffet; reset 19 Aug 2026)
  • Treat grok.com Build /loop (7-day) as Grok Bot production scheduling
  • Treat Heavy four-at-once as named desks with routines

Definitive article

A Skill Is How. A Routine Is When.

Other skills: ai-search-visibility · always-reply · boil-the-ocean · client-access-checklist · client-relationship-cadence · content-agent

Public URL is /skills-skill-then-routine/, never /skills/skill-then-routine/. The rule: canonical skill rule.

The full run order is on the skill-pack directory. Every skill here is one task from the Task Library — the library is the catalogue of what can be done; a pack is the subset you install; an agent is who runs it.

Where this sits in the system

Context is what the worker knows. A skill is the method it follows. A pack is a folder of methods. None of them does any work alone. Work happens when a job runs the right skill against the right context on a schedule, checks its output, and leaves a receipt the next worker can inspect.

  1. Context — the verified facts, goals, evidence, decisions, and working state kept in user-owned Markdown, Obsidian, Drive, or Git so any authorized model can pick up the same work.
  2. Skill — one task, written down to a standard, so an agent can run it without you in the room. There are 243 of them.
  3. PackYOU ARE HERE — those skills bundled into a download you install in one paste.
  4. Agent — a named role with a job description — not a chat window you retype every morning.
  5. Job — a schedule, a QA cycle, and somewhere to keep working files. Miss any of the three and nothing runs twice.
  6. Proof — every finished run written up in public, and the lesson pushed back into the skill.

The map: The System · every asset: Asset Tracker · next door: Asset Tracker.

Scroll to Top