The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill incorporates AI-powered threat analysis which processes external security data.\n
Ingestion points: The k7-console --analyze-threat command and the analyze_with_ai function in the Bash script take threat-data as input. This data, such as malicious file names or logs, is potentially attacker-controlled.\n
Boundary markers: The instructions do not specify any boundary markers (like delimiters or instructions to ignore embedded commands) when passing threat data to the AI model.\n
Capability inventory: The skill possesses high-privilege capabilities, including the ability to isolate endpoints (k7-console --isolate) and trigger automated responses.\n
Sanitization: There is no evidence of data sanitization or validation performed on the threat data before it is processed by the AI integration.\n- [EXTERNAL_DOWNLOADS]: The skill's installation procedures involve downloading components from external URLs.\n
The Windows installation script fetches an agent installer from https://k7-releases.example/K7-Agent-16.0.1195-x64.msi.\n
The Linux configuration script downloads an APT public key from https://apt.k7security.example/pubkey.gpg.\n- [COMMAND_EXECUTION]: The skill frequently executes system-level commands through subprocess calls and shell scripts to perform agent management, scans, and policy enforcement.\n
The Python script uses subprocess.run to call the k7-console management tool.\n
The Bash script executes k7-console, curl, and jq for threat monitoring and alerting.\n- [PRIVILEGE_ESCALATION]: Deployment and configuration tasks require elevated system permissions.\n
The installation scripts for Windows, macOS, and Linux utilize administrative or root access (e.g., msiexec, sudo, sudo apt-key).

k7-total-security-unified-defense