The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill downloads and executes scripts and binaries from external GitHub repositories. Evidence includes fetching the LinPEAS script from the peass-ng repository and piping it directly to the shell in SKILL.md, and downloading the PwnKit exploit binary from a public repository in references/advanced-privesc.md.
[COMMAND_EXECUTION]: The skill utilizes a wide range of powerful shell commands to discover and exploit system vulnerabilities. This includes patterns for sudo abuse, SUID exploitation, and capability-based privilege escalation, as well as instructions for modifying sensitive system configurations like /etc/passwd and cron jobs.
[DATA_EXFILTRATION]: The skill instructs the agent to harvest sensitive information from the target host, such as SSH private keys, database credentials, and application environment variables. It also searches through shell history files and process environment blocks for plaintext secrets.
[PRIVILEGE_ESCALATION]: The primary objective of the skill is to gain elevated permissions on a Linux host. It details methods for abusing SUID/SGID binaries, misconfigured cron jobs, and specific Linux capabilities, and outlines techniques for exploiting kernel vulnerabilities like DirtyCow and DirtyPipe.
[INDIRECT_PROMPT_INJECTION]: The skill demonstrates a vulnerability surface by ingesting untrusted data that could contain malicious instructions. 1. Ingestion points: Reads external content from shell history, log files, and configuration files (documented in references/credential-harvest.md). 2. Boundary markers: No delimiters or warnings are used to prevent the agent from following instructions embedded in the ingested data. 3. Capability inventory: The skill has broad capabilities including file writing, network downloads, and root-level command execution. 4. Sanitization: There is no evidence of input validation or sanitization before processing file contents.

post-exploit-linux