add-skill
Pass
Audited by Gen Agent Trust Hub on May 2, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches configuration and documentation from official Claude and Anthropic resources (code.claude.com and github.com/anthropics/skills).
- [COMMAND_EXECUTION]: Generates shell commands to create directory structures for new skills. It includes specific validation rules (lowercase letters, numbers, and hyphens) for user-provided names to prevent command injection.
- [PROMPT_INJECTION]: The skill processes external documentation from the web to guide file generation, creating a surface for indirect prompt injection. This is mitigated by targeting trusted official domains and providing clear templates for the generated content.
- Ingestion points: SKILL.md (via WebFetch and WebSearch tool calls)
- Boundary markers: Absent
- Capability inventory: mkdir, Write tool, bash execution
- Sanitization: Absent for web content; strict allow-list validation for user-provided names.
Audit Metadata