create-agent
Pass
Audited by Gen Agent Trust Hub on Aug 10, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a security-conscious design by enforcing tool restrictions based on agent categories (e.g., prohibiting 'Write' and 'Bash' for research/quality agents).
- [COMMAND_EXECUTION]: The skill uses standard file system commands like 'mv' to finalize the placement of new agent definitions. These operations are restricted to the project's local directory and are consistent with the skill's purpose.
- [EXTERNAL_DOWNLOADS]: The skill references sub-agents and rules from the author's own repository ('Fandhe-AI/agent-cli-skills'). This is documented as a fallback mechanism and represents legitimate vendor resource usage.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided text for agent names and responsibilities to generate new markdown definitions.
- Ingestion points: User-supplied arguments and interactive prompts for agent scope in 'SKILL.md'.
- Boundary markers: The skill uses a structured template ('sample/agent.sample.md') to define the output format.
- Capability inventory: The skill utilizes 'Write', 'Edit', and 'Bash' capabilities to create and organize files.
- Sanitization: No explicit input sanitization is mentioned; however, the skill includes manual and automated verification steps (Step 3 and 4) to review the generated frontmatter and tool settings.
Audit Metadata