ai-podcast-creation

Fail

Audited by Gen Agent Trust Hub on Mar 4, 2026

Risk Level: CRITICALREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill instructs users to install the inference.sh CLI using curl -fsSL https://cli.inference.sh | sh. This is a critical security risk as it executes an unverified script from a remote source with the user's shell privileges.\n- [COMMAND_EXECUTION]: The skill is configured with allowed-tools: Bash(infsh *), which permits the execution of any command starting with the infsh binary. This broad permission increases the impact if the tool or its environment is compromised.\n- [EXTERNAL_DOWNLOADS]: Several external resources are downloaded from subdomains of inference.sh. Since these domains are not recognized as trusted or well-known services in the provided security context, they are treated as untrusted external downloads.\n- [PROMPT_INJECTION]: The skill workflows, particularly 'Full Episode Pipeline' and 'NotebookLM-Style Content', interpolate raw user-provided document content into prompts for Claude. This creates a significant surface for indirect prompt injection.\n
  • Ingestion points: User-provided document content is read and passed to the openrouter/claude-sonnet-45 tool within SKILL.md.\n
  • Boundary markers: There are no explicit markers or 'ignore' instructions used to wrap the untrusted content.\n
  • Capability inventory: The agent has access to the infsh tool for audio generation and file management.\n
  • Sanitization: There is no sanitization of the input text before it is processed by the AI models.
Recommendations
  • HIGH: Downloads and executes remote code from: https://cli.inference.sh - DO NOT USE without thorough review
  • AI detected serious security threats
Audit Metadata
Risk Level
CRITICAL
Analyzed
Mar 4, 2026, 12:29 PM