User-typed slash commands do not show "Successfully loaded skill" indicator

Description

When a user manually types a slash command (e.g., /spec, /commit, /do), the skill content loads correctly but the visual indicator does not appear in the terminal.

The indicator only appears when the model calls the Skill tool programmatically during its response. User-typed slash commands use a direct injection path that bypasses the Skill tool entirely.

Steps to Reproduce

1. Create a project with custom skills in .claude/skills/
2. Type any slash command, e.g. /spec
3. Observe: skill content loads into the conversation, but no "Successfully loaded skill" indicator appears
4. Compare: when the model calls Skill("spec") during its response (e.g., via a PostToolUse hook), the indicator DOES appear

Expected Behavior

User-typed slash commands should show the same "Successfully loaded skill" visual indicator as model-initiated Skill tool calls.

Actual Behavior

• User-typed: content loads silently via direct injection, no indicator
• Model-initiated: Skill(skill-name) indicator with "Successfully loaded skill" appears

Additional Context

• The PreToolUse[Skill] hook also doesn't fire for user-typed slash commands, since the Skill tool is not called
• This means any skill gating logic in hooks is bypassed for user-typed invocations
• The skill content itself loads correctly in both paths — this is purely about the visual indicator and hook execution