Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@tools/modules/system/runner-cleanup/runner-cleanup`:
- Line 43: The case arm handling -c|--config) currently dereferences $2
directly; guard against a missing argument by checking that the next positional
exists before assigning to CONFIG: inside the -c|--config) branch test whether
"${2:-}" (or $# >= 2) is set, and if not call the script's usage/failure path
and exit with a non-zero status, otherwise set CONFIG="$2" and shift 2; update
the -c|--config) branch accordingly to avoid an unbound-variable error under set
-u.

In `@tools/modules/system/runner-cleanup/runner-cleanup.service`:
- Line 3: The Documentation= line currently points to a misspelled or incorrect
URL; update the value of the Documentation= field so it points to the correct
Armbian configng repository URL (replace the existing
"https://github.com/armbian/configng" value with the verified, correctly spelled
repository URL) and confirm the link resolves in a browser; look for and edit
the line beginning with "Documentation=" to make this change.

---

Nitpick comments:
In `@tools/modules/system/runner-cleanup/README.md`:
- Around line 63-75: Two fenced code blocks in the README containing the
snippets "runner-cleanup: no actions-runner-* users on this host — nothing to
do." and the multiline example starting with "wipe
/home/actions-runner-01/_work/" lack language identifiers and trigger MD040; fix
by adding a language tag (use text for plain output and sh for shell commands)
to those fenced blocks so they become e.g. ```text and ```sh respectively to
satisfy the markdown linter.

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@tools/modules/system/module_armbian_runners.sh`:
- Line 171: The script currently silences failures of "systemctl enable --now
runner-cleanup.timer", which can hide a failed timer install; change this to run
the command without redirecting output, capture its exit code, and if non-zero
log a clear error (including command stderr) and attempt a safe fallback: try
"systemctl enable runner-cleanup.timer" then "systemctl start
runner-cleanup.timer", checking and logging each exit status; reference the
invocation "systemctl enable --now runner-cleanup.timer" in the fix and ensure
any error messages are emitted instead of using ">/dev/null 2>&1 || true".

In `@tools/modules/system/runner-cleanup/README.md`:
- Line 113: The README has fenced code blocks missing language identifiers;
update the two blocks containing the messages "runner-cleanup: no
actions-runner-* users on this host — nothing to do." and the block starting
"wipe /home/actions-runner-01/_work/" to use a language tag (e.g., change the
opening ``` to ```text) so markdown linters correctly detect the code block
language.

In `@tools/modules/system/runner-cleanup/runner-cleanup`:
- Around line 143-145: The early exit when runner_users is empty (the if block
checking ${`#runner_users`[@]} and calling note "runner-cleanup: no
actions-runner-* users..." then exit 0) prevents the later housekeeping steps
(section 4: journald/APT/log cleanup) from running; change the flow so that when
runner_users is empty you still perform the housekeeping: remove or relocate the
exit 0 and ensure the script continues to execute section 4 after logging the
"no actions-runner-* users" note (leave the note call intact but do not return
early), or alternatively call the housekeeping function/section explicitly
before exiting.
- Around line 231-237: The block that currently wipes "${work}" when "$unit" is
empty is unsafe; instead of running "find \"$work\" -mindepth 1 -delete" when no
systemd unit is found, change it to skip the wipe and emit a clear log like the
existing "skipping wipe to avoid mid-job damage" message (referencing the
variables unit, work, and user) so the script avoids deleting _work if the
runner cannot be quiesced; keep the early "log" message but replace the
destructive "run find ... -delete" with a skip-and-continue path that mirrors
the safe behavior used elsewhere.
- Around line 253-258: The current check using command -v docker only verifies
the CLI exists but not the daemon; update the logic so that before attempting
any container/image cleanup you try a daemon call (e.g., docker info) and if
that call fails treat Docker as unavailable: set goto_housekeeping and log a
skip rather than letting the script abort under set -euo pipefail. Apply the
same pattern around the docker cleanup blocks that call docker container prune
and docker image prune so those commands are executed only when the daemon check
succeeds; reference and update the conditional that currently uses command -v
docker, the goto_housekeeping variable, and the docker container prune / docker
image prune invocations.
- Around line 241-244: The cleanup currently exits if run find "$work" -mindepth
1 -delete fails (due to set -euo pipefail), preventing the later run systemctl
start "${unit}.service" from executing; change the flow so the restart is always
attempted: run systemctl stop "${unit}.service", then perform the wipe but do
not let failures abort the script (e.g. run the find with its failure masked or
capture its exit status), record any wipe error and continue, and finally always
attempt run systemctl start "${unit}.service" and call note "WARNING: failed to
restart ${unit}.service..." only if the restart itself fails. Ensure you
reference and modify the existing run invocation for find "$work" -mindepth 1
-delete and the subsequent run systemctl start "${unit}.service" and note call.

In `@tools/modules/system/runner-cleanup/runner-cleanup.timer`:
- Line 2: The timer unit runner-cleanup.timer has Description="Daily
runner-cleanup" but its OnCalendar=hourly schedule; update the Description in
runner-cleanup.timer (the Description field) to reflect the actual schedule
(e.g., "Hourly runner-cleanup" or "Runs hourly: runner-cleanup") so the
systemctl list-timers output is accurate; ensure any other similar unit with the
same mismatch (noted as 19-19) is updated the same way.