automations git:20261023.611d35bA

Immutable. This exact content never changes and is served at /api/v1/blob/d5de037a1c2e38a7.

---
name: automations
description: "Schedule or manage recurring and one-shot HYPERFIDE agent or script automations."
---

# Automations

An automation is a scheduled task. When due it runs in one of two modes:

agent Spawn a thread or re-prompt a target thread with a configured prompt.
script Run a stored server-side script and capture stdout/stderr/exit.

Use the top-level `Hyperfide automation` command. The CLI routes it to this plugin.

Pass `--project` explicitly for every automation command. Inside a thread, automations are stamped origin `agent` and record the creating thread automatically. Automation-spawned threads cannot create automations.

Personal supports automations with `--project proj_personal`. Use `hyperfide-platform project list --include-personal --json` to include it in discovery; an empty default project list does not mean Personal is unavailable.

Choosing a mode:

Use `script` when the output is fully determined by code: watchdogs, threshold alerts, health checks, heartbeats, and API pollers with a fixed output shape. Scripts run on the hyperfide-platform server. New standard-project scripts use the server-host project source when one exists; Personal and projects without one run in the plugin's shared script storage directory. Existing scripts without a saved policy also run there. Use `--working-directory automation-storage|project|<absolute-server-path>` to select the policy. Script automations do not have an environment field and do not accept environment flags.

Design the script to print nothing when there is nothing to report: an exit-0 run with empty stdout/stderr, or a last non-empty line of `{"wakeAgent": false}`, is recorded as a skipped silent tick. Any other output is captured; non-zero exit or timeout is recorded as a failed run.

Use `agent` when the run needs reasoning: summarize a feed, pick interesting items, draft a human-friendly message, or branch on content.

Creating:

```bash
Hyperfide automation create --project <id> --name "..." [schedule flags] [mode flags]
```

For creation flags and mode-specific defaults, read
[references/creation.md](references/creation.md).

Read `references/script-runtime.md` before you use a script file, depend on
injected variables, or diagnose retries, timeouts, restarts, and silent runs.

Managing:

```bash
Hyperfide automation list --project <id>
Hyperfide automation list --target-thread <threadId>
Hyperfide automation show <automationId> --project <id>
Hyperfide automation update <automationId> --project <id> [--name <name>] [schedule flags] [complete execution flags | partial agent update flags]
Hyperfide automation pause <automationId> --project <id>
Hyperfide automation resume <automationId> --project <id>
Hyperfide automation run <automationId> --project <id> [--idempotency-key <key>]
Hyperfide automation runs <automationId> --project <id> [--limit <count>] [--output <runId>]
Hyperfide automation delete <automationId> --project <id> --yes
```

`list --target-thread` returns automations in any project that re-prompt that
thread; the thread's Info tab shows the same list. It does not include
automations the thread merely created.

For partial updates, mode replacement, execution targets, or damaged records,
read [references/updates.md](references/updates.md). Every command supports `--json`.

`Hyperfide automation <command> --help` prints that command's exact arguments, options,
accepted values, and rules. Unknown commands, unknown options, and stray
arguments are rejected rather than ignored, and a failure under `--json` also
prints a `{"ok":false,"error":{"code","message","hint"}}` envelope on stdout.