2.13.0-beta.20). Features may change before the next stable release.Switch to stable →ak activity
ak activity list
Read a bounded newest-first snapshot from the local AgentKit activity log and consume its exact JSON event shape.
Use ak activity list when you need a finite snapshot of recent local
skill-run events. The command reads the local append-only log and does not wait
for new events.
Usage
ak activity listThe command accepts no positional arguments.
Command flags
| Flag | Default | Description |
|---|---|---|
--limit <n> | 100 | Return at most the most recent n matching events. Values at or below zero currently fall back to 100. |
--since <id> | Empty | Keep events whose IDs are lexicographically greater than this cursor. |
Shared flags are described in CLI conventions.
Read recent events
ak activity list --limit 20
ak activity list --since 00000000000000000000-000000 --limit 20 --jsonFiltering happens before the limit. The command keeps events after --since,
takes the newest matching entries up to --limit, then returns them newest
first. --since is an opaque cursor comparison; copy an exact id from an
earlier result rather than inventing one.
Missing logs produce an empty successful result. Empty or malformed NDJSON
lines are skipped, so total counts only decoded events returned by this
invocation.
Human output
Human output prints a summary followed by each event's id, kind, skill,
and timestamp. It does not display the optional runtime, duration, or exit code;
use JSON when those fields matter.
JSON output
Success writes one envelope to stdout:
schema_version: 1
kind: activity.list
data.schema_version: 1
data.total: integer
data.events: arrayEach event can contain:
id, ts, kind, skill, runtime, duration_ms, exit_codekind is currently run.started, run.completed, or run.failed. Fields
marked optional by the event schema are omitted when empty or zero. Gate on the
outer schema_version and kind before reading data.
Data and retention boundary
The command reads ~/.agentkit/activity/events.ndjson; AGENTKIT_HOME
overrides the base. It does not create, rewrite, delete, or prune the log. The
--limit and --since flags only bound this response and do not change data on
disk.
No network, authentication, entitlement, provider, subprocess, TTY, or stdin
behavior is involved. --yes has no effect.
Exit behavior
| Exit | Meaning |
|---|---|
0 | The snapshot was returned, including an empty snapshot. |
1 | The local log could not be read, output failed, or a positional argument was supplied. |
2 | A flag value could not be parsed. |
In --json mode, command runtime and argument failures can produce no JSON
error body. Flag-parse failures print plain-text usage to stderr.
On macOS, the CLI-wide first-launch guard can also write
~/.agentkit/.first-run-darwin and print a one-time hint to stderr before this
command runs.