Please address the comments from this code review:

## Overall Comments
- The parsing of `avail` values assumes a strictly numeric prefix (`sed -n 's/.*avail=\([0-9]*\).*/\1/p'`), but `df -BGB` typically emits values with unit suffixes (e.g. `13G`/`13GB`); consider normalizing by stripping all non-digits (e.g. `tr -dc '0-9'`) once and reusing that to avoid brittle parsing.
- When `first_avail < min_avail` (e.g. if disk is freed over time), `peak_consumed=$((first_avail - min_avail))` becomes negative; it may be clearer to clamp this to zero or take the absolute/maximum difference to avoid confusing negative "consumption" values in the summary.

## Individual Comments

### Comment 1
<location path=".github/actions/job-preamble/action.yaml" line_range="146-148" />
<code_context>
+      with:
+        shell: bash
+        run: |
+          LOGFILE="/tmp/disk-usage-monitor.log"
+          PIDFILE="/tmp/disk-usage-monitor.pid"
+
+          # Record initial snapshot
</code_context>
<issue_to_address>
**suggestion (bug_risk):** Use per-job unique paths for LOGFILE/PIDFILE to avoid cross-job interference on shared runners

On self-hosted or reused runners, multiple jobs can share /tmp and overlap in time. Fixed LOGFILE/PIDFILE names can cause jobs to interfere with each other (e.g., killing another job’s monitor or overwriting its log). Please derive these paths from something unique like $GITHUB_RUN_ID, $GITHUB_JOB, and/or $$ (e.g., /tmp/disk-usage-monitor-${GITHUB_RUN_ID}-${GITHUB_JOB}.log).

```suggestion
        run: |
          # Use per-job unique paths to avoid cross-job interference on shared runners
          LOGFILE="/tmp/disk-usage-monitor-${GITHUB_RUN_ID:-unknown-run}-${GITHUB_JOB:-unknown-job}-$$.log"
          PIDFILE="/tmp/disk-usage-monitor-${GITHUB_RUN_ID:-unknown-run}-${GITHUB_JOB:-unknown-job}-$$.pid"
```
</issue_to_address>

### Comment 2
<location path=".github/actions/job-preamble/action.yaml" line_range="203-210" />
<code_context>
+          min_ts=""
+          first_avail=""
+          last_avail=""
+          for entry in "${entries[@]}"; do
+            avail_val=$(echo "$entry" | sed -n 's/.*avail=\([0-9]*\).*/\1/p')
+            entry_ts=$(echo "$entry" | cut -d' ' -f1)
+            if [[ -z "$first_avail" ]]; then
+              first_avail="$avail_val"
+            fi
+            last_avail="$avail_val"
+            if [[ "$avail_val" -lt "$min_avail" ]]; then
+              min_avail="$avail_val"
+              min_ts="$entry_ts"
</code_context>
<issue_to_address>
**issue (bug_risk):** Guard against empty or unparsable avail values before doing integer comparisons

If df or the log format ever produce a non-integer or missing avail value, sed will leave avail_val empty and [[ "$avail_val" -lt "$min_avail" ]] will raise a non-integer operand error, which can break this step. Consider validating avail_val with something like [[ "$avail_val" =~ ^[0-9]+$ ]] and skipping or handling entries that don’t match before assigning first_avail/last_avail or doing -lt comparisons.
</issue_to_address>