The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data from external web applications, creating a potential attack surface for indirect prompt injection attacks.
Ingestion points: Data enters the agent's context through curl output, whatweb fingerprinting, browser crawling (browser_navigate), and inspection of files like robots.txt and sitemap.xml during Phase 2 (Recon).
Boundary markers: The instructions do not specify the use of clear delimiters or instructions for the agent to disregard potential commands or instructions embedded within the retrieved target content.
Capability inventory: The skill allows the agent to execute network tools (nmap, curl), write files to the local system (findings, logs, evidence), and perform multi-step exploitation tasks based on analyzed data.
Sanitization: The skill recommends redacting captured tokens and credentials to the last 6 characters before logging them in chat history, but it does not specify sanitization procedures for the arbitrary HTML or text content analyzed during vulnerability discovery.
[COMMAND_EXECUTION]: The skill relies on executing terminal-based networking and reconnaissance tools to fulfill its primary purpose.
Evidence: The scripts/recon-scan.sh wrapper script executes nmap, curl, and whatweb. The agent is also instructed to use curl for tech fingerprinting and header analysis throughout the pentesting workflow.
[EXTERNAL_DOWNLOADS]: The skill performs active network requests to external URLs provided by the user for security testing purposes.
Evidence: The skill uses curl and a browser tool to interact with target web applications. These network operations are constrained by an authorization gate and a scope allowlist to ensure requests are directed only at intended targets.
[DYNAMIC_EXECUTION]: The skill uses dynamically generated script snippets to perform utility functions during the reconnaissance phase.
Evidence: scripts/recon-scan.sh utilizes a Python heredoc to perform IP-in-CIDR checks, ensuring that resolved IP addresses for target hostnames remain within the authorized scope.

web-pentest