The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous shell commands for network discovery, service enumeration, and API probing. It utilizes tools such as nmap, curl, grpcurl, openssl, and the GitHub CLI (gh).
[INDIRECT_PROMPT_INJECTION]: The shell commands incorporate variables provided by the user, such as $TARGET, $CONTROLLED_USER_A_ID, and $TARGET_ORG. This creates an attack surface where malicious input could attempt command injection if the agent does not properly sanitize these inputs.
Ingestion points: User-provided variables in SKILL.md (e.g., $TARGET, $CONTROLLED_USER_A_ID, $CONTROLLED_USER_B_ID, $TARGET_ORG).
Boundary markers: Includes a 'Mandatory confirmation gate' that instructs the agent to ask for target details and wait for explicit confirmation before executing commands.
Capability inventory: Executes subprocesses via nmap, curl, grpcurl, openssl, gh, and tee.
Sanitization: The shell command templates do not include explicit logic for sanitizing or escaping the user-provided variables.
[EXTERNAL_DOWNLOADS]: The skill references external tools and repositories such as grpcurl, grpcui, and the elementalsouls/Claude-BugHunter repository on GitHub for additional security testing capabilities.

hunt-grpc