How to author and inspect durable schedules from agent tools
Scanned 9/2/2026
Install to Claude Code
npx -y skills add lukacf/meerkat --skill schedule-workflow --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Schedule Workflow?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/lukacf-schedule-workflow)More formats (shields.io, HTML) on the badges page.
---
name: schedule-workflow
description: How to author and inspect durable schedules from agent tools
requires_capabilities: [schedule]
---
# Schedule Workflow
Use Schedule when something should happen later, repeatedly, or on a wall-clock
calendar. Schedules own time and delivery. They do not own live work state.
## Operating Rules
- Create schedules with `meerkat_schedule_create` when the user asks for a
reminder, recurrence, follow-up, monitor, wakeup, or routine automation.
- Choose the smallest trigger that matches the request: `once` for one future
instant, `interval` for fixed cadence, `calendar` for wall-clock recurrence
in a named timezone.
- For the current agent session, use the agent-facing session target
`{"target_kind":"session","type":"current_session","action":{"type":"prompt","prompt":"Check in"}}`.
The host resolves it to a persisted durable target.
- Use `resumable_session` only when you have the exact durable `session_id`,
and `materialize_on_demand_session` when the schedule must create a session
from an explicit build spec on first fire.
- The defaults are `misfire_policy: {"type":"skip"}`,
`overlap_policy: "skip_if_running"`, and
`missing_target_policy: "mark_misfired"`. Change them only when the user asks
for bounded catch-up, concurrency, or different missing-target behavior.
- Inspect with `meerkat_schedule_get`, `meerkat_schedule_list`, and
`meerkat_schedule_occurrences` before creating duplicates.
- Pause for temporary suspension, resume to reactivate, update to change future
behavior, and delete only when the schedule should stop permanently. Delete
is irreversible, but historical occurrences remain queryable.
## Boundaries
- Use WorkGraph for pending work, dependencies, claims, and evidence.
- Use Schedule for time. A scheduled prompt may ask an agent to inspect
WorkGraph, but Schedule should not duplicate WorkGraph readiness logic.
- Treat an occurrence claim and delivery outcome as Schedule authority only.
They do not authorize a WorkGraph claim or prove that the underlying shared
work completed.
- Do not infer successful delivery from a due time. Inspect occurrence phase;
pending, claimed, dispatching, awaiting completion, completed, skipped,
misfired, superseded, and delivery failed are distinct states.
- Use memory for recalled knowledge, not for future wakeups.
- Use builtin tasks for private scratch items that do not need scheduled
delivery or shared durable coordination.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!