[DOCS] Skills docs should document interactive-tool exception to `allowed-tools` auto-approval behavior

Documentation Type

Unclear/confusing documentation

Documentation Location

https://code.claude.com/docs/en/skills

Section/Topic

allowed-tools behavior for interactive tools (for example AskUserQuestion) when a skill is active.

Current Documentation

The skills docs currently state:

Skills that define allowed-tools grant Claude access to those tools without per-use approval when the skill is active.

What's Wrong or Missing?

Changelog v2.1.69 includes a bug fix indicating interactive tools were being silently auto-allowed when listed in a skill's allowed-tools, bypassing permission prompts and returning empty answers.

The current documentation does not call out this interactive-tool edge case, so readers may assume all tools listed in allowed-tools always bypass per-use approval.

Suggested Improvement

Add an explicit note in the allowed-tools section clarifying interactive-tool behavior:

• Interactive tools (for example AskUserQuestion) should not be silently auto-approved in a way that bypasses required user interaction.
• If interactive tools are allowed in skills, documentation should explain how permission prompts and user input collection behave.
• Include one concrete example showing expected behavior when AskUserQuestion appears in a skill workflow.

Impact

Medium - Makes feature difficult to understand

Additional Context

Affected Pages:

| Page | Context |
|------|---------|
| https://code.claude.com/docs/en/skills | Primary description of allowed-tools behavior in skills |
| https://code.claude.com/docs/en/settings | Tool reference and permission model context (AskUserQuestion, permission semantics) |
| https://platform.claude.com/docs/en/agent-sdk/skills | Agent SDK cross-reference |

Total scope: 3 pages affected

Source: Changelog v2.1.69

Exact changelog entry:

Fixed interactive tools (e.g., AskUserQuestion) being silently auto-allowed when listed in a skill's allowed-tools, bypassing the permission prompt and running with empty answers