[DOCS] GitHub Actions Migration Guide recommends destructive `--system-prompt` flag instead of `--append-system-prompt`

Documentation Type

Missing documentation (feature not documented)

Documentation Location

https://code.claude.com/docs/en/github-actions

Section/Topic

The "Upgrading from Beta" section, specifically the "Breaking Changes Reference" table.

Current Documentation

In the Breaking Changes Reference table, the documentation currently maps the legacy input to the new CLI argument as follows:

| Old Beta Input | New v1.0 Input |
| :--- | :--- |
| custom_instructions | claude_args: --system-prompt |

What's Wrong or Missing?

The migration guide instructs users to convert custom_instructions (which were typically used to add context like coding standards) to the --system-prompt CLI flag.

However, according to the CLI Reference, the --system-prompt flag replaces the entire system prompt, which removes all default Claude Code instructions and tool definitions. Using this flag effectively wipes Claude Code's ability to function as an agent or use tools.

The intention of custom_instructions in the beta action was to provide additional context, not to wipe the agent's core personality. Therefore, recommending --system-prompt will break workflows for users following this guide.

Suggested Improvement

The table should be updated to recommend --append-system-prompt, which preserves Claude Code's built-in capabilities while adding the custom instructions.

Suggested change:

| Old Beta Input | New v1.0 Input |
| :--- | :--- |
| custom_instructions | claude_args: --append-system-prompt |

Impact

High - Prevents users from using a feature

Additional Context

• Reference for flag behavior: https://code.claude.com/docs/en/cli-reference
• Quote: "--system-prompt: Use when you need complete control over Claude's system prompt. This removes all default Claude Code instructions, giving you a blank slate."
• Quote: "--append-system-prompt: Use when you want to add specific instructions while keeping Claude Code's default capabilities intact. This is the safest option for most use cases."