claude-agent-sdk-agent-creation
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [Command Execution] (LOW): The skill grants the agent access to the Bash tool via the allowed-tools configuration. This provides a broad command execution surface that requires restricted environments.
- [Indirect Prompt Injection] (LOW): The skill is susceptible to Indirect Prompt Injection through the use of settingSources which ingests content from external files. 1. Ingestion points: .claude/CLAUDE.md and ~/.claude/CLAUDE.md referenced in configuration snippets. 2. Boundary markers: Absent; instructions are loaded directly into the agent context. 3. Capability inventory: Bash, Write, Edit, Read, Grep, Glob. 4. Sanitization: None provided in the configuration templates.
Audit Metadata