The Agent Skills Directory

[COMMAND_EXECUTION]: The execute_command(command) function defined in references/tools.md allows the AI agent to execute arbitrary shell commands on the host system. This is a high-risk capability that can be misused to bypass other tool constraints or access sensitive system resources.
[DYNAMIC_EXECUTION]: The skill includes several functions designed to generate and test malicious code at runtime. Specifically, ai_generate_payload, advanced_payload_generation, and generate_exploit_from_cve in references/tools.md facilitate the creation of attack code tailored to a target's context, including built-in evasion techniques.
[REMOTE_CODE_EXECUTION]: By automating the transition from external threat intelligence (CVE metadata) to active exploit generation through generate_exploit_from_cve, the skill significantly lowers the barrier for automated code execution attacks against target infrastructure.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill actively ingests untrusted data from a wide array of external sources including httpx_probe (web headers/content), katana_crawl (crawled web data), gau_discovery (URL history), nuclei_scan (vulnerability reports), and monitor_cve_feeds (external threat data).
Boundary markers: The instructions lack delimiters or explicit warnings to ignore instructions embedded within the data retrieved by these tools.
Capability inventory: The skill possesses powerful capabilities including arbitrary shell execution (execute_command), SQL injection (sqlmap_scan), and automated payload generation.
Sanitization: There is no evidence of sanitization or validation logic for the external content before it is processed by the agent or used to influence subsequent attack steps.

pentest-osint-recon