diff --git a/CLAUDE.md b/CLAUDE.md
index f5d2e24..2f09f8c 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -32,3 +32,13 @@ custom-claude-skills/
 - **Restrict tools** with `allowed-tools` to the minimum needed — this reduces permission prompts
 - **Use $ARGUMENTS** for user input, `$0`, `$1` etc. for positional args
 - Dynamic commands in `!`command`` run at skill load time, not during Claude's execution
+
+## `!`command`` Gotchas
+
+The `!`command`` syntax in SKILL.md runs shell commands at skill load time to pre-gather context. These commands go through Claude Code's Bash permission checker against the `allowed-tools` patterns. Key constraints:
+
+- **No `$()` command substitution** — the permission checker rejects commands containing `$()` even if the outer command matches an allowed pattern. This means you cannot nest subshells (e.g., `git log --since="$(git log ...)"` will fail).
+- **No complex shell pipelines relying on subshells** — keep `!`command`` commands simple and self-contained. If you need complex logic, have the skill instructions tell Claude to run it via tool calls instead.
+- **`allowed-tools` patterns must match the command binary** — `Bash(git *)` allows any `git` subcommand, but `Bash(cat *)`, `Bash(sed *)`, `Bash(ls *)` must be listed separately if those commands appear in `!`command`` blocks.
+- **Prefer specific tool patterns over broad ones** — `Bash(git log *)`, `Bash(git diff *)` is safer than `Bash(git *)` to avoid accidentally allowing destructive git commands (e.g., `git reset`, `git push --force`).
+- **Fallback to tool instructions for dynamic paths** — if a command needs `$ARGUMENTS` to compute a file path (e.g., `cat scripts/verify-m${N}.sh`), replace the `!`command`` with a plain-text instruction telling Claude to use the Read tool. This avoids `$()` substitution and is more reliable.
diff --git a/skills/reflect/SKILL.md b/skills/reflect/SKILL.md
index af3a9fe..107055d 100644
--- a/skills/reflect/SKILL.md
+++ b/skills/reflect/SKILL.md
@@ -4,7 +4,7 @@ description: >
   Run a milestone reflection after completing a milestone. Reviews the full conversation history,
   git log, and current project docs to produce structured reflection artifacts. Use when a milestone
   is complete and verified. Invoke with the milestone number, e.g. /reflect M8
-allowed-tools: Read, Edit, Write, Grep, Glob, Bash(git *)
+allowed-tools: Read, Edit, Write, Grep, Glob, Bash(git log *), Bash(git diff *), Bash(git show *), Bash(cat *), Bash(sed *), Bash(ls *)
 ---
 
 # Milestone Reflection Skill
@@ -19,8 +19,8 @@ detour, every wrong assumption, every backtrack — and distill it into structur
 
 ## Pre-gathered context
 
-### Git log for this milestone
-!`git log --oneline --since="$(git log --all --oneline --grep="[Mm]$(echo '$ARGUMENTS' | tr -d 'Mm' | awk '{print $1-1}').*[Vv]erif\|[Mm]$(echo '$ARGUMENTS' | tr -d 'Mm' | awk '{print $1-1}').*[Rr]eflect\|[Mm]$(echo '$ARGUMENTS' | tr -d 'Mm' | awk '{print $1-1}').*[Cc]omplete" --format='%aI' | head -1 || echo '3 months ago')" 2>/dev/null || git log --oneline -30`
+### Git log (last 50 commits)
+!`git log --oneline -50`
 
 ### Current MEMORY.md
 !`cat MEMORY.md 2>/dev/null || echo "No MEMORY.md found"`
@@ -35,7 +35,7 @@ detour, every wrong assumption, every backtrack — and distill it into structur
 !`ls -1 scripts/ 2>/dev/null || echo "No scripts directory"`
 
 ### Verify script for this milestone
-!`cat scripts/verify-m$(echo '$ARGUMENTS' | tr -d 'Mm' | awk '{print $1}').sh 2>/dev/null || echo "No verify script found for this milestone"`
+(Read the verify script manually using the Read tool: `scripts/verify-m<N>.sh` where N is the milestone number from $ARGUMENTS)
 
 ## Instructions