agent-creator
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided descriptions and behavioral prompts to generate agent configuration files.
- Ingestion points: User input gathered in Steps 1 through 5 of SKILL.md, including system prompts and regex patterns for tool settings.
- Boundary markers: Absent; user input is directly interpolated into the profile template.
- Capability inventory: The skill uses file-writing capabilities to create config.json and Markdown rule files in the local directory ~/.aider-desk/agents/.
- Sanitization: No validation or sanitization logic is described for user-supplied strings, posing a potential risk if malicious input is provided for sensitive fields like bash tool regex patterns.
- [COMMAND_EXECUTION]: The skill performs file system operations to manage agent profiles.
- Evidence: Step 8 in SKILL.md directs the agent to create directories and write configuration files to the local file system.
- [DYNAMIC_EXECUTION]: The skill generates functional configuration files at runtime based on user input and pre-defined templates.
- Evidence: The skill uses assets/templates/config.json.template to assemble structured JSON profiles that dictate the runtime permissions and behavior of subsequent agent instances.
Audit Metadata