atlas-paperclipforge-agent-creator
Warn
Audited by Gen Agent Trust Hub on Jun 17, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONCREDENTIALS_UNSAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill provides SQL templates and instructions for the agent to execute data manipulation queries (INSERT, DELETE) against the 'agents' and 'agent_models' tables. This allows the agent to modify the system's operational database based on incoming requests.
- [CREDENTIALS_UNSAFE]: The documentation includes templates for 'ADD-agent-api-keys.sql', explicitly directing the agent to manage authentication secrets and API access configurations within the database.
- [INDIRECT_PROMPT_INJECTION]: The 'Agent Specification' and 'Skill Definition' workflows involve taking potentially untrusted natural language descriptions of agent roles and interpolating them into SQL statements and YAML configurations. This presents an attack surface where a malicious specification could attempt SQL injection or configuration manipulation if the executing environment does not apply strict sanitization.
- [PRIVILEGE_ESCALATION]: The skill provides examples of assigning high-level permissions, such as 'database_admin: true', to newly created agents. This capability allows for the creation of new entities with elevated privileges across the ecosystem.
Audit Metadata