Keep in-repo agent scratch out of the build context and out of git (closes #27)

The canonical .dockerignore and .gitignore both omitted the in-repo agent
scratch directory, which holds one worktree per in-flight agent, so under
`COPY . .` an entire extra checkout of the repo reached the image. The
two entries are deliberately different shapes: anchored in .dockerignore,
where the `**/` form would also delete a legitimately named nested
directory, and unanchored in .gitignore, where a pattern already matches
at every depth. Anchoring leaves a gap where agents run in
subdirectories, stated in the vendored file itself. The second half is
the consequence of excluding .git: `git describe` in a build stage yields
an empty version without erroring, so the version is now computed on the
host and passed in.

Model: opus-5

script/docker

@@ -10,7 +10,16 @@ ROOT="$(cd "$SCRIPT_DIR/.." && pwd -P)"
 
 main() {
     cd "$ROOT"
-    docker build --no-cache -t "$("$SCRIPT_DIR/projectname")" .
+    # Own line: a failing command substitution inside an argument does
+    # not trip `set -e`, so the inline form degrades silently to an
+    # empty constant. VERSION is computed here because .dockerignore
+    # excludes .git, so `git describe` in a build stage yields an empty
+    # version without failing.
+    version="$(git describe --tags --always --dirty 2>/dev/null || true)"
+    [ -n "$version" ] || version="unknown"
+    docker build --no-cache \
+        --build-arg VERSION="$version" \
+        -t "$("$SCRIPT_DIR/projectname")" .
 }
 
 main "$@"