oracle
Pass
Audited by Gen Agent Trust Hub on Mar 9, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill metadata defines the installation of the
@steipete/oraclepackage from the npm registry as a required dependency. - [COMMAND_EXECUTION]: The skill provides instructions for executing the
oraclecommand-line tool and usingnpx -yto run commands from the npm registry without manual confirmation. - [DATA_EXFILTRATION]: The documented tool is designed to transmit local file contents and prompts to external AI model providers (OpenAI, Gemini) or via browser automation.
- [PROMPT_INJECTION]: The skill exposes an indirect prompt injection surface by ingesting repository file content via glob patterns for processing by an LLM.
- Ingestion points: Local files selected through
--filearguments in theoracleCLI as described inSKILL.md. - Boundary markers: No specific boundary markers or instructions are provided to the model to ignore embedded instructions in the ingested file content.
- Capability inventory: The
oraclebinary can read local files and perform network requests to external API endpoints or browser engines. - Sanitization: No explicit sanitization or validation of the file content is mentioned before it is sent to the external model.
Audit Metadata