The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads and installs several security utilities (subfinder, httpx, nuclei) from the official GitHub repositories of ProjectDiscovery, which is a well-known provider of security tooling.
[COMMAND_EXECUTION]: The instructions execute multiple command-line tools for reconnaissance and scanning, including subfinder, httpx, nuclei, and amass. These are standard operations for the stated purpose of attack surface management.
[CREDENTIALS_UNSAFE]: The documentation includes placeholders for Shodan and Censys API keys (e.g., 'YOUR_SHODAN_API_KEY'). No actual secrets are hardcoded in the provided file.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection (Category 8) because it ingests untrusted data from external reconnaissance (subdomains, service headers, and vulnerability scan results) and uses it as input for subsequent tool executions. However, the risk is minimal as the processing is performed by specialized security tools rather than directly by an LLM interpreting the data as instructions.
Ingestion points: subdomains.txt, combined_subdomains.txt, previous_scan.json.
Boundary markers: None present in the command-line piping logic.
Capability inventory: Shell execution of subfinder, httpx, nuclei, and a local agent.py script.
Sanitization: No explicit sanitization or validation of the discovered subdomains or service data is described before they are passed to other tools.

implementing-attack-surface-management