The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The Dockerfile downloads numerous security tools and utilities from external sources including GitHub (ProjectDiscovery, OWASP, Assetnote, tomnomnom), PortSwigger, and the Oh My Zsh project. These are recognized as well-known or trusted sources in the security community.
[COMMAND_EXECUTION]: The skill includes an execute_command tool defined in references/tools.md. This tool allows the agent to run arbitrary shell commands within the container environment, intended as a fallback for tools not explicitly mapped to specific functions.
[PRIVILEGE_ESCALATION]: The compose.yaml file specifies network_mode: host, which grants the container full access to the host machine's network interfaces. While typical for network scanning tools (like nmap or naabu), this bypasses standard container network isolation.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data from external web applications during security assessments, creating a potential surface for indirect prompt injection.
Ingestion points: Untrusted web content, API responses, and crawler output gathered by tools such as katana_crawl, httpx_probe, and nuclei_scan (referenced in SKILL.md and references/workflows.md).
Boundary markers: None identified in the provided templates to distinguish between instructions and ingested data.
Capability inventory: High-privilege capabilities including arbitrary command execution (execute_command), extensive network access (network_mode: host), and file system access for report generation.
Sanitization: No explicit evidence of sanitization or validation of external tool outputs before they are processed by the agent.

seclens-enterprise-web