The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Surface.
Ingestion points: The script scripts/agent.py ingests data from external sources via enumerate_subdomains_crtsh, enumerate_dns_records, shodan_org_search, check_web_technologies, and search_github_leaks.
Boundary markers: No delimiters or explicit instructions are provided to the agent to ignore or isolate potential malicious instructions embedded in the gathered OSINT data.
Capability inventory: The skill possesses network request capabilities and filesystem write access.
Sanitization: No sanitization or validation of the data retrieved from external APIs or target headers is performed before output generation.
[COMMAND_EXECUTION]: The SKILL.md workflow documentation provides the agent with several shell commands for external tools like amass, subfinder, dig, and wafw00f. While these are standard security tools, their execution involves interacting with external environments based on target-provided strings.
[SAFE]: The skill interacts with well-known and reputable services including Google DNS, GitHub, Shodan, and crt.sh for legitimate information gathering purposes.
[SAFE]: The check_web_technologies function in scripts/agent.py disables SSL verification (verify=False). This is a documented best-practice violation but is common in security reconnaissance tools to maintain compatibility with varied target configurations.

conducting-external-reconnaissance-with-osint