The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a comprehensive suite of shell commands using curl and bash loops designed to perform active network reconnaissance and exploitation against user-specified targets.
[DATA_EXFILTRATION]: Includes specific patterns for exfiltrating sensitive system data from remote targets, such as reading /proc/self/environ to harvest environment variables which may contain AWS keys or other service credentials.
[REMOTE_CODE_EXECUTION]: Contains ready-to-use exploit payloads for triggering Remote Code Execution on vulnerable Node.js backends via Prototype Pollution (targeting child_process and lodash sinks) and Server-Side Template Injection (SSTI) in engines like EJS, Pug, and Handlebars.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill is designed to ingest and process data from external targets (e.g., checking if specific keys like polluted or isAdmin appear in API responses).
Boundary markers: None. The skill does not implement delimiters or instructions to ignore potential commands embedded in target responses.
Capability inventory: The skill has access to shell execution and network operations, creating a high-risk path if a target returns malicious instructions.
Sanitization: No sanitization or validation of external data is performed before the agent processes the response.
[REMOTE_CODE_EXECUTION]: Although the automated scanner flagged a command as being 'piped to node', manual inspection reveals this was a false positive on a grep command; however, the skill still contains numerous RCE exploit payloads intended for delivery to remote systems.

hunt-nodejs