Support for Claude Code skills (slash commands) in automation workflows

## Question

Is there a way to invoke Claude Code skills (defined in `.claude/skills/`) from automation workflows?

## Context

I have skills defined in my repo:

```
.claude/skills/
├── syner-daily-standup -> ../apps/dev/skills/syner-daily-standup
└── ...
```

Each skill is a markdown file with frontmatter defining `allowed-tools` and instructions.

## What I tried

```yaml
- uses: anthropics/claude-code-action@v1
with:
prompt: "/syner-daily-standup"
claude_args: |
--agent syner
--max-turns 10
```

Claude exits in ~44ms with 1 turn and $0 cost, suggesting the skill isn't being loaded.

## Expected behavior

I expected `/syner-daily-standup` to be recognized and executed as a skill, similar to how it works in the interactive Claude Code CLI.

## Questions

1. Are skills supported in claude-code-action?
2. If not, is there a workaround (e.g., reading the skill file and passing its content as prompt)?
3. I noticed `INPUT_EXPERIMENTAL_SLASH_COMMANDS_DIR` in `action.yml` - is this related?

Thanks!

Read action.yml, especially INPUT_EXPERIMENTAL_SLASH_COMMANDS_DIR, then trace the workflow entry point handling the prompt and claude_args shown in the issue. Verify whether .claude/skills/ and /syner-daily-standup are loaded by this action; done should be a documented supported invocation or a confirmed workaround.

Written by the indexing model from the issue text.