The Agent Skills Directory

[DATA_EXFILTRATION]: The skill documents a payload designed to exfiltrate Kubernetes service account tokens to an external domain (attacker.com) using the PostgreSQL COPY FROM PROGRAM command combined with curl.
[REMOTE_CODE_EXECUTION]: The skill provides multiple methods to achieve remote code execution via SQL injection, including writing PHP web shells to the web root using MySQL's INTO OUTFILE, executing shell commands via MSSQL's xp_cmdshell, utilizing PostgreSQL's COPY FROM PROGRAM functionality, and running Java commands through Oracle's DBMS_JAVA.RUNJAVA.
[COMMAND_EXECUTION]: The automation workflow instructs the agent to execute a variety of command-line tools for reconnaissance and exploitation, such as sublist3r, httpx, waybackurls, sqlmap, ghauri, hakrawler, and arjun.
[CREDENTIALS_UNSAFE]: Instructions are provided for accessing sensitive cloud metadata endpoints (AWS and Azure IMDS) to retrieve security credentials, and for poisoning database connection pools in serverless environments to escalate privileges.
[OBFUSCATION]: The skill includes a dedicated section on WAF bypass techniques, instructing the agent on how to use URL encoding, hex encoding, and comment injection to hide malicious payloads from security parsers.
[INDIRECT_PROMPT_INJECTION]: The skill describes an attack surface where instructions are embedded within common web data formats like JSON, GraphQL, and WebSocket messages, which could influence the behavior of agents processing these data streams.
Ingestion points: Processes URL parameters, POST bodies, cookies, headers, and API filter objects.
Boundary markers: None specified.
Capability inventory: Includes shell command execution and file write capabilities via SQLi payloads.
Sanitization: Focused on bypassing sanitization rather than implementing it.
[EXTERNAL_DOWNLOADS]: The skill references a methodology checklist hosted on the author's GitHub repository (github.com/SnailSploit/offensive-checklist).

offensive-sqli