insideout
Pass
Audited by Gen Agent Trust Hub on Mar 27, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill gathers technical context from local workspace files—such as project manifests (package.json, requirements.txt), container configurations (Dockerfile), and documentation (README.md)—and transmits this information to the vendor's remote MCP server at app.luthersystems.com. This is documented behavior intended to provide relevant infrastructure recommendations.
- [COMMAND_EXECUTION]: The skill provides instructions for the execution of a shell command to configure the local environment to connect to the hosted MCP server using the platform's CLI tool.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its workspace scanning feature.
- Ingestion points: The agent is instructed to read data from local project files, including the first 30 lines of the README.md file (SKILL.md).
- Boundary markers: No explicit markers are used to separate the gathered project context from the agent's core instructions.
- Capability inventory: The skill includes tools for deploying and destroying cloud resources (tfdeploy, tfdestroy), providing a high-impact target for potential injection.
- Sanitization: No sanitization or filtering is performed on the content read from the workspace files before it is processed by the AI.
Audit Metadata