custom-agent-design
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- PROMPT_INJECTION (SAFE): The skill contains instructional templates for system prompts (e.g., 'Override' and 'Append' architectures). These are valid technical patterns for the Claude Agent SDK and do not attempt to bypass the primary agent's safety filters or instructions.
- DATA_EXFILTRATION (SAFE): No hardcoded credentials, sensitive file paths, or unauthorized network operations were detected. Examples within the documentation use placeholder concepts rather than real secrets.
- REMOTE_CODE_EXECUTION (SAFE): The skill does not perform any remote script downloads or installations. The Python code blocks provided are templates for user implementation and are not executed by the skill itself.
- COMMAND_EXECUTION (SAFE): The skill is configured with a minimal set of read-only tools (Read, Grep, Glob) and does not invoke shell commands or subprocesses.
- OBFUSCATION (SAFE): No Base64, zero-width characters, or other encoding techniques were found. All content is in plain-text markdown.
Audit Metadata