The Agent Skills Directory

[COMMAND_EXECUTION]: The helper script assets/nmap_scan_template.sh is vulnerable to command injection. It constructs a shell command by concatenating several variables derived from user-supplied arguments (such as the target hostname and port specifications) and then executes the resulting string using the eval command. An attacker could provide a malicious target string (e.g., 127.0.0.1; whoami) or a crafted option to execute arbitrary shell commands with the privileges of the user running the script.
[PROMPT_INJECTION]: The skill's workflow presents an indirect prompt injection surface because it is designed to ingest and analyze untrusted external data.
Ingestion points: The agent is instructed to read application source code (via Grep), dependency audit reports (npm audit, pip-audit), and SAST tool outputs (Semgrep, Gitleaks).
Boundary markers: There are no delimiters or instructions provided to isolate this untrusted external content from the agent's core instructions.
Capability inventory: The agent has powerful capabilities, including Bash execution, Write, and Edit file system access, which could be abused if the agent is manipulated by instructions hidden in the data it analyzes.
Sanitization: No sanitization, validation, or filtering is performed on the data retrieved from external files or tool outputs before it is processed by the agent.

performing-security-testing