nest agents under their dispatching skill

Top-level agents/ was an arbitrary flat layout from the skeleton
commit, with no substantive reason behind it. Mirroring AILang's
convention (skills/<name>/agents/<agent>.md) is strictly better:
it makes the "no orphan agents" rule structurally true — you
cannot place an agent file outside a skill — and it keeps the
skill/agent binding visible at first glance.

Claude Code's user-level discovery is unaffected: install.sh now
symlinks ~/.claude/agents/<name> to skills/<name>/agents/ for
each skill that has an agents/ subdirectory. The .claude/
registry stays flat; the source tree stays nested. The two
checklists previously living in agents/README.md are folded
into skills/README.md.

install.sh

@@ -1,11 +1,13 @@
 #!/usr/bin/env bash
 # Install the skills plugin globally into ~/.claude/.
 #
-# Creates symlinks from ~/.claude/skills/<name> and
-# ~/.claude/agents/<name> into this repo's skills/ and agents/
-# subdirectories. Idempotent — existing symlinks pointing to
-# this repo are left alone; existing entries that point elsewhere
-# are reported and skipped.
+# For each skill <name> under skills/<name>/, creates symlinks:
+#   ~/.claude/skills/<name>  -> <repo>/skills/<name>
+#   ~/.claude/agents/<name>  -> <repo>/skills/<name>/agents   (if that dir exists)
+#
+# Idempotent — existing symlinks pointing to this repo are left
+# alone; existing entries that point elsewhere are reported and
+# skipped.
 
 set -euo pipefail
 
@@ -42,14 +44,9 @@ if [ -d "$REPO_DIR/skills" ]; then
         [ -d "$s" ] || continue
         name="$(basename "$s")"
         link_one "$s" "$CLAUDE_SKILLS/$name"
-    done
-fi
-
-if [ -d "$REPO_DIR/agents" ]; then
-    for a in "$REPO_DIR"/agents/*/; do
-        [ -d "$a" ] || continue
-        name="$(basename "$a")"
-        link_one "$a" "$CLAUDE_AGENTS/$name"
+        if [ -d "$s/agents" ]; then
+            link_one "$s/agents" "$CLAUDE_AGENTS/$name"
+        fi
     done
 fi