intelligems-morning-brief
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- Persistence Mechanisms (HIGH): The skill establishes persistence on the host system by creating a macOS LaunchAgent in Step 5.4. This allows scripts to run automatically without user intervention.\n
- Evidence: SKILL.md Step 5.4 calls
setup_automation.shwhich is described as creating a macOS LaunchAgent.\n- Indirect Prompt Injection (MEDIUM): The skill prompts the user for a Slack Webhook URL and interpolates it directly into shell commands. This creates a vulnerability where malicious input could lead to command injection.\n - Ingestion points: User prompt for Slack Webhook URL in Step 5.\n
- Boundary markers: Absent. The input is wrapped in double quotes in the script but not sanitized.\n
- Capability inventory: Shell execution (
bash,python3), file system modification (cp,echo).\n - Sanitization: Absent.\n- Data Exposure (LOW): The skill stores and reads the
INTELLIGEMS_API_KEYin a plaintext.envfile. While local, this exposes credentials to any process with read access to the directory.\n - Evidence: SKILL.md Step 1.
Recommendations
- AI detected serious security threats
Audit Metadata