aiconfig-create
Pass
Audited by Gen Agent Trust Hub on Apr 29, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill instructs the agent to process untrusted user input and interpolate it into remote configurations, creating a surface for indirect prompt injection. 1. Ingestion points: User-provided use cases, instructions, and messages (SKILL.md). 2. Boundary markers: The skill does not provide delimiters or instructions to isolate user-provided text from the configuration payload structure. 3. Capability inventory: The skill uses the setup-ai-config and create-ai-config-variation MCP tools to write this data to the LaunchDarkly platform. 4. Sanitization: There is no instruction to validate or sanitize the content of instructions or messages before they are sent to the remote API.
- [PROMPT_INJECTION]: The 'Bias Towards Action' section directs the agent to 'proceed through the entire workflow without stopping' and 'complete all steps (create + verify) in one pass.' This instruction encourages high autonomy and reduces opportunities for user review during multi-step remote operations.
Audit Metadata