AGENT LAB: SKILLS

ai-podcast-creation

Fail

Audited by Gen Agent Trust Hub on Feb 19, 2026

Risk Level: CRITICALREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • REMOTE_CODE_EXECUTION (CRITICAL): The skill instructs the user or agent to execute curl -fsSL https://cli.inference.sh | sh. This pattern downloads and executes a shell script directly from the internet without prior verification, allowing for full system compromise.\n- EXTERNAL_DOWNLOADS (HIGH): The skill relies on binaries and scripts hosted on inference.sh, which is not among the predefined trusted organizations. Under the [TRUST-SCOPE-RULE], this source is unverified and represents a significant supply chain risk.\n- COMMAND_EXECUTION (MEDIUM): The skill uses the Bash(infsh *) tool, granting the agent broad permission to execute any subcommand of the infsh utility, which includes network operations and file manipulation.\n- PROMPT_INJECTION (LOW): The 'NotebookLM-Style Content' section creates an Indirect Prompt Injection surface. Evidence: 1. Ingestion point: <your-document-content> in SKILL.md. 2. Boundary markers: Absent. 3. Capability inventory: Subprocess calls via infsh and file-write operations. 4. Sanitization: None. Instructions embedded in processed documents could influence agent behavior.
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
Feb 19, 2026, 03:41 AM