poc-validator - SKILL.md Agent Skill

name: poc-validator description: Proof-of-Concept (PoC) Validation Skill for security exploit verification. Use when validating exploitability of vulnerabilities, generating tailored payloads, executing exploits in sandboxed environments, or verifying successful exploitation (e.g., SQL injection, CVE exploitation, command injection). Triggers on tasks involving payload generation, exploit execution verification, sandboxed testing, or confirmation of administrative access/data exfiltration after exploitation attempts.

PoC-Validator Skill

Validates vulnerability exploitability through controlled payload generation, sandboxed execution, and success verification.

Quick Start

Generate and validate a PoC for SQL injection:

from scripts.payload_generator import PayloadGenerator
from scripts.sandbox_executor import SandboxExecutor
from scripts.verify_exploit import ExploitVerifier

# Generate payload
gen = PayloadGenerator()
payload = gen.generate_sql_payload(
    target_url="http://target.com/login",
    parameter="username",
    technique="error_based"
)

# Execute in sandbox
executor = SandboxExecutor()
result = await executor.execute_sqlmap(
    target="http://target.com/login",
    payload=payload,
    risk_level=2
)

# Verify success
verifier = ExploitVerifier()
is_exploitable = await verifier.verify_sql_injection(result)

Workflow

1. Payload Generation

Use scripts/payload_generator.py for:

CVE-based exploits: Lookup known payloads from vulnerability_mapping.json
SQL Injection: Error-based, Union-based, Time-based, Boolean-based
Command Injection: System command payloads with evasion techniques
Path Traversal: Directory traversal sequences
XXE/XML Injection: External entity payloads

See references/payload_patterns.md for complete payload library.

2. Sandbox Execution

Use scripts/sandbox_executor.py for isolated exploit testing:

Docker container isolation with resource limits
Network isolation (no egress to private ranges)
Execution timeouts (default: 300s)
Read-only filesystem with tmpfs for output
Automatic cleanup post-execution

Critical: All exploit execution MUST use sandbox. Never run exploits directly on host.

3. Success Verification

Use scripts/verify_exploit.py to confirm exploitation:

SQL Injection: Detect database version extraction, table enumeration
RCE: Verify command output in response
Auth Bypass: Confirm session token/admin access
Data Exfiltration: Detect sensitive data patterns in responses
File Upload: Verify file presence/execution

Risk Levels

Level	Payload Type	Tools	Use Case
0	Passive detection	None	Vulnerability confirmation only
1	Safe payloads	SQLMap --risk 1	Non-destructive testing
2	Active exploitation	SQLMap --risk 2, Metasploit	Controlled exploitation
3	Aggressive	All tools with full payloads	Full PoC validation

Integration with Zen-AI-Pentest

# Integration with ExploitAgent
from autonomous.exploit_validator import ExploitValidator

def validate_poc(self, vulnerability: dict) -> dict:
    """Validate vulnerability with PoC execution."""
    validator = ExploitValidator(
        sandbox_enabled=True,
        risk_level=vulnerability.get("risk", 1)
    )

    return validator.validate(
        vuln_type=vulnerability["type"],
        target=vulnerability["target"],
        evidence=vulnerability["evidence"]
    )

Safety Controls

Before any exploit execution:

Validate target is not in private ranges (guardrails/ip_validator.py)
Confirm explicit authorization in database
Set execution timeout
Enable Docker sandbox
Log all actions to audit trail

See references/safety_requirements.md for complete safety checklist.

Error Handling

Error	Cause	Solution
Sandbox timeout	Payload too slow	Increase timeout or use faster technique
Container crash	Resource exhaustion	Reduce threads, increase memory limit
Network blocked	Egress filtering	Verify allowed outbound in config
Verification failed	WAF/IPS blocking	Try evasion techniques from payload_patterns.md

Output Format

{
  "poc_id": "uuid",
  "vulnerability_id": "vuln-123",
  "status": "verified|failed|inconclusive",
  "exploit_type": "sql_injection|rce|xxe|...",
  "payload_used": "...",
  "sandbox_logs": "...",
  "evidence": {
    "before": "...",
    "after": "...",
    "extracted_data": "..."
  },
  "risk_level": 2,
  "timestamp": "2026-03-20T14:46:00Z",
  "verified_by": "exploit_agent_v3"
}

References

Payload Patterns Library - Complete payload collection
Sandbox Setup Guide - Docker configuration details
Safety Requirements - Mandatory safety checks
CVE Database Schema - CVE lookup structure

Scripts

scripts/payload_generator.py - Generate tailored exploit payloads
scripts/sandbox_executor.py - Execute exploits in Docker sandbox
scripts/verify_exploit.py - Verify exploitation success
scripts/cleanup_sandbox.py - Clean up sandbox resources