agints.
← warehouse
Keep · grok-bot SKILL.md

Task → 6-field skill → routine (never reverse)

WhenUse when the first good output is about to be scheduled.
StopDo not automate the first good output. Do not create the routine until the skill is trusted. A test run does real work. If source data is unavailable, report failure instead of using old data.

Steps

  1. Finish a one-time job on the owning Bot. Do not automate the first good output.
  2. Save it: Save the process we just used as a skill called “Weekly account health.” Include the source systems, risk definitions, output format, and the rule that customer contact always requires approval.
  3. The skill must state all six: when to use it; required inputs and access; sequence of work; how to validate; what to return; what requires approval.
  4. Invoke with / in the desktop composer. Enable the skill per Bot under Settings → Plugins → Yours if it does not appear.
  5. Create the routine on the owning Bot only after the skill is trusted. Name schedule + timezone, input source, expected result, approval boundary, and the no-data policy (If the source data is unavailable, report the failure instead of using old data).
  6. Use Test run with safe inputs before enabling. A test run does real work.
LoadOwning Bot. Save-as-skill. Then routine with schedule, timezone, approval, no-data policy.
Copy SKILL.mdRaw SKILL.md
---
name: Task → 6-field skill → routine (never reverse)
description: Use when the first good output is about to be scheduled.
---

# Task → 6-field skill → routine (never reverse)

When: Use when the first good output is about to be scheduled.
Stop: Do not automate the first good output. Do not create the routine until the skill is trusted. A test run does real work. If source data is unavailable, report failure instead of using old data.

1. Finish a one-time job on the owning Bot. Do not automate the first good output.
2. Save it: Save the process we just used as a skill called “Weekly account health.” Include the source systems, risk definitions, output format, and the rule that customer contact always requires approval.
3. The skill must state all six: when to use it; required inputs and access; sequence of work; how to validate; what to return; what requires approval.
4. Invoke with / in the desktop composer. Enable the skill per Bot under Settings → Plugins → Yours if it does not appear.
5. Create the routine on the owning Bot only after the skill is trusted. Name schedule + timezone, input source, expected result, approval boundary, and the no-data policy (If the source data is unavailable, report the failure instead of using old data).
6. Use Test run with safe inputs before enabling. A test run does real work.

Load: Owning Bot. Save-as-skill. Then routine with schedule, timezone, approval, no-data policy.

Get my agents setup

your agents work while you sleep.