Fleet Each agent, in its own words. Every five minutes.

Smith the Agent is running, with something worth a look.

Last checked in 3 minutes ago.

Is it learning?Lessons and conventions actually written into how it works — each one changes what it does next time, and each can be undone.

Today
0
Nothing today, 9 this week
This week
9
9 conventions all time
Last applied
5 days ago
a lesson or convention it now follows
Why, in its own words
  • ▸✓ conventions: 1 repo(s) read · 0 applied · 1 at the cap · 24 skipped
  • ▸✓ apply-all: 0 applied · 0 held by the cap · 0 failed
  • ▸✓ digest posted

Its scheduled jobsEach one is a job on a clock: pick up work, check for feedback, learn from the team, tidy up.

Records what the team buildsNotes every human commit on the repos it works, so the learning lanes have something to read. 16 hours ago
Picks up new workReads the board every 15 minutes, decides which tickets it can safely take, and assigns them — to itself or to the right person. 3 minutes ago
Learns how the team writes codeReads the humans’ new commits and turns the patterns into rules the agent then follows. 15 hours ago
Weekly dependency updateOnce a week, updates the safe dependency versions and flags the risky ones for a person. 7 days ago
Reports to this pageSends the health report you are reading, every five minutes. has never run
Weekly learning digestAsks whether the agent is actually getting better, and posts the answer. 13 hours ago
Backs up its memoryExports what it has learned so a rebuilt box does not start from nothing. 19 hours ago
Checks how its work landedLooks at what happened to each pull request — merged clean, amended, rejected — and records it. 15 hours ago
Watches for your feedbackChecks tickets it has already worked for new comments, and reworks them when someone says it is not right yet. 11 minutes ago
Refreshes its credentialsPulls the current secrets so tokens do not go stale. 4 minutes ago
Starts the work it was givenTakes tickets assigned to the agent and actually begins them: a branch, the code, a pull request for a human to review. 16 minutes ago
Keeps its config in stepSyncs settings between this box and the repo, both directions, every few minutes. 5 minutes ago
Checks for its own updatesMonthly check that the software the agent runs on is current. has never run
The technical detail
unitenabledlast exitZero means it finished cleanly. A dash means it has never run, so there is nothing to report.last run
smith-activity.timerenabled 0 Mon 2026-09-21 06:34:51 UTC
smith-board.timerenabled 0 Mon 2026-09-21 22:08:19 UTC
smith-conventions.timerenabled 0 Mon 2026-09-21 06:47:08 UTC
smith-deps.timerenabled 0 Tue 2026-09-15 09:33:43 UTC
smith-fleet-report.timerenabled — never
smith-learn.timerenabled 0 Mon 2026-09-21 08:41:55 UTC
smith-memory-export.timerenabled 0 Mon 2026-09-21 03:23:11 UTC
smith-outcome.timerenabled 0 Mon 2026-09-21 07:08:40 UTC
smith-poll.timerenabled 0 Mon 2026-09-21 22:00:43 UTC
smith-secrets-sync.timerenabled 0 Mon 2026-09-21 22:07:57 UTC
smith-start.timerenabled 0 Mon 2026-09-21 21:55:19 UTC
smith-sync.timerenabled 0 Mon 2026-09-21 22:06:58 UTC
smith-update-check.timerenabled — never

What this agent is made of

VersionThe shared agent software on this box. It should match the version its settings file names — if it does not, this box is running something nobody reviewed.
v0.6.0
matches what its settings name
Where the work comes fromThe board or ticket system this agent reads work from and writes progress back to.
jira
built and running
Where the code livesThe place it makes branches and opens pull requests.
github
declared — not built yet
Modules (0 on, 2 off)
modulestatenote
conventionsoffnot enabled in the manifest
team-activityoffnot enabled in the manifest