Skill File Validator Incorrectly Flags Valid Attributes

Bug Report: Skill File Validator Incorrectly Flags Valid Attributes

Environment

• Claude Code Version: 2.1.32
• Package: @anthropic-ai/claude-code@2.1.32
• Platform: macOS (Darwin 24.4.0)

Description

The built-in skill file validator incorrectly reports user-invocable and allowed-tools as unsupported attributes, even though they are documented as valid in the official Claude Code documentation.

Steps to Reproduce

1. Create a skill file with the following frontmatter:

---
name: test-skill
description: Test skill
argument-hint: test
allowed-tools: Bash
user-invocable: true
---

1. Load the skill in Claude Code
2. Observe linter warnings in system diagnostics

Expected Behavior

No warnings should be shown for allowed-tools and user-invocable as they are documented valid attributes.

Actual Behavior

The validator shows these warnings:

⚠ [Line 5:1] Attribute 'allowed-tools' is not supported in skill files. Supported: argument-hint, compatibility, description, disable-model-invocation, license, metadata, name, user-invokable.
⚠ [Line 6:1] Attribute 'user-invocable' is not supported in skill files. Supported: argument-hint, compatibility, description, disable-model-invocation, license, metadata, name, user-invokable.

Issue Analysis

The warnings are contradictory - they claim the attributes are "not supported" but then list them in the "Supported" list. This suggests either:

1. The validator's supported attributes list is outdated/incorrect
2. There's a bug in the validation logic itself

Documentation Reference

According to the official documentation at https://code.claude.com/docs/en/slash-commands#skill-tool:

• user-invocable (with 'c') is a valid skill attribute
• allowed-tools is a valid skill attribute

Both attributes function correctly despite the warnings.

Impact

• Confusing for skill developers
• False positive warnings clutter the diagnostic output
• May cause developers to remove valid attributes

Suggested Fix

Update the skill file validator to match the official documentation and remove false positive warnings for user-invocable and allowed-tools.

Additional Context

• The skill functions correctly with these attributes
• Only affects the linter/validator, not runtime functionality