The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface for indirect prompt injection by design.
Ingestion points: The agent is instructed to execute DAST tools and AI-driven curl probes, ingesting the resulting tool output and HTTP responses for analysis (SKILL.md, references/implementation.md).
Boundary markers: No specific boundary markers or instructions are provided to help the agent distinguish between tool metadata and potentially malicious instructions embedded in a target application's HTTP response headers or body.
Capability inventory: The skill grants the agent broad shell command execution and network probing capabilities to interact with external systems.
Sanitization: The skill does not define methods for sanitizing or validating external tool output before the agent processes it.
[COMMAND_EXECUTION]: The skill provides specific shell commands for invoking various security auditing tools, including ZAP-CLI, Nuclei, Nikto, and ffuf. These tools perform active network scanning and interact with external endpoints.
[DATA_EXFILTRATION]: The skill provides instructions for manual curl probes targeting sensitive server-side files such as .env, .git/config, and internal metadata services (SSRF probes). While these are intended for security discovery, they involve requesting sensitive configuration data over the network.

common-dast-tooling