rag-cli-operator
Pass
Audited by Gen Agent Trust Hub on Apr 24, 2026
Risk Level: SAFECREDENTIALS_UNSAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill provides instructions to store sensitive API keys (e.g., POE_API_KEY, OPENAI_API_KEY) as machine-level environment variables via PowerShell. This practice exposes credentials to all users and processes on the operating system and encourages unnecessary privilege escalation.
- [COMMAND_EXECUTION]: The skill relies on shell command execution via the
uvtool to perform workspace initialization, document ingestion, and querying. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through its document processing workflow.
- Ingestion points: External document content is ingested from local paths provided via the
--srcargument inSKILL.md. - Boundary markers: The skill does not implement or mention the use of delimiters or specific instructions to isolate untrusted document content from agent instructions.
- Capability inventory: The skill allows for command execution (
uv run rag-cli) and querying of the ingested data using an LLM. - Sanitization: No sanitization, validation, or filtering of the content within the ingested documents is documented or implemented.
Audit Metadata