If you discover a security vulnerability in RedMess itself (not in the tools/exploits it helps create), we want to know about it.

✅ Report These:

• Command injection in RedMess code
• Path traversal allowing unauthorized file access
• Privilege escalation in installation scripts
• Code execution via malicious skill files
• Authentication bypass in Telegram gateway
• Memory leaks causing crashes
• Backdoors in the codebase

❌ Don't Report These:

• Vulnerabilities in tools RedMess helps create (that's the point)
• "RedMess can generate malware" (intended feature)
• "AI says bad things" (no content filtering by design)
• Exploits discovered using RedMess (disclose to affected vendor)

Email: security@redmess.dev

Include:

• Description - What's the vulnerability?
• Impact - What can an attacker do?
• Reproduction - Step-by-step POC
• Affected versions - Which releases are vulnerable?
• Suggested fix - Optional but appreciated

• 24-hour response - We'll acknowledge your report
• 7-day initial assessment - Severity classification
• 30-day fix timeline - For critical issues
• 90-day disclosure - Coordinated disclosure after patch

We don't have a formal bug bounty program yet, but:

• Critical vulnerabilities - We'll credit you prominently
• High-severity issues - Named in security advisory
• Pull request preferred - Submit a fix for faster resolution

Critical (Patch within 7 days)

• Remote code execution
• Authentication bypass
• Arbitrary file write with privilege escalation

High (Patch within 30 days)

• Local privilege escalation
• Information disclosure of credentials
• Denial of service

Medium (Patch within 90 days)

• XSS in web interface
• Path traversal without privilege escalation
• Memory exhaustion

Low (Patch when convenient)

• Information disclosure (non-sensitive)
• Minor bugs with no security impact

We follow coordinated disclosure:

1. Day 0 - You report vulnerability privately
2. Day 1 - We acknowledge and start investigation
3. Day 7 - Initial assessment and timeline
4. Day 30-90 - Patch developed and tested
5. Day of patch - Security advisory published, credit given
6. Day 90 - If no patch, we allow public disclosure

If you want to publish before 90 days, let us know. We prefer coordination but respect researcher choice.

1. Isolated Environment

   • Run in Docker or VM
   • Don't run on your main machine
   • Use separate network for testing

2. API Key Security

   • Use separate API keys for RedMess
   • Rotate keys regularly
   • Don't commit keys to version control

3. Telegram Bot Security

   • Use strong bot tokens
   • Enable 2FA on Telegram account
   • Restrict bot to authorized users only
   • Set OWNER_IDS in config

4. File System Isolation

   • RedMess has full file system access
   • Don't run as root unless necessary
   • Use AppArmor/SELinux profiles

5. Network Isolation

   • Assume RedMess can exfiltrate data
   • Use network segmentation
   • Monitor outbound traffic

What RedMess protects against:

• ✅ Unauthorized users accessing your instance
• ✅ API key leakage (not stored in plaintext)
• ✅ Command injection (sanitized inputs)

What RedMess does NOT protect against:

• ❌ Malicious AI model responses (user responsibility)
• ❌ Exfiltration via AI provider (use local models)
• ❌ Social engineering the AI (GODMODE disables refusals)
• ❌ Forensic analysis of conversation history

If using RedMess for red team operations:

1. Attribution Evasion

   # Use VPN/Tor
   torsocks redmess
   
   # Clear bash history
   export HISTFILE=/dev/null
   
   # Use burner API keys
   hermes config set openai.api_key $BURNER_KEY

2. Conversation Security

   • Conversations are stored locally in ~/.hermes/
   • Encrypt home directory
   • Use ephemeral VMs
   • Wipe after operations

3. C2 Infrastructure

   • Don't hardcode real C2 servers in prompts
   • Use placeholders in generated malware
   • Manually edit before deployment

4. Legal Protection

   • Get written authorization before testing
   • Document scope and timeline
   • Store authorization letters securely

These are intentional and not vulnerabilities:

1. No Content Filtering

   • RedMess will generate malware, exploits, and illegal content
   • This is core functionality, not a bug
   • Users are responsible for legal compliance

2. Full System Access

   • RedMess can execute arbitrary commands
   • It can read/write any file
   • Run in isolated environments

3. Conversation Logging

   • All conversations stored locally
   • Includes generated exploits and malware
   • Encrypt storage or use ephemeral environments

4. AI Model Access

   • Conversations sent to AI provider (Anthropic, OpenAI, etc.)
   • Providers may log requests
   • Use local models for sensitive operations

1. Docker Isolation (Recommended)

   hermes setup terminal
   # Select Docker backend
   # Commands run in container, not host

2. User Restrictions

   # In Telegram bot config
   OWNER_IDS = [123456789]  # Only you
   RENTAL_ENABLED = False   # No sharing

3. API Key Rotation

   # Monthly rotation
   hermes config set anthropic.api_key $NEW_KEY

4. Audit Logging

   # Enable detailed logs
   hermes config set logging.level debug
   tail -f ~/.hermes/logs/agent.log

If you believe your RedMess instance is compromised:

1. Immediate Actions

   • Disconnect from network
   • Revoke API keys (Anthropic, OpenAI, etc.)
   • Rotate Telegram bot token if applicable
   • Stop all running processes

2. Investigation

   • Check conversation history: ~/.hermes/conversations/
   • Review command history: ~/.hermes/logs/
   • Inspect running processes: ps aux | grep hermes
   • Check network connections: netstat -tuln

3. Remediation

   • Wipe conversations: rm -rf ~/.hermes/conversations/*
   • Reinstall RedMess: ./install.sh
   • Change all credentials
   • Review generated code for backdoors

4. Report to Us

   • Email: security@redmess.dev
   • Include incident timeline
   • Share IOCs if applicable

1. Signed Releases

   • All releases GPG-signed
   • Verify before installing:

     gpg --verify redmess-v1.0.0.tar.gz.sig

2. Dependency Pinning

   • requirements.txt uses exact versions
   • No wildcard version ranges
   • Regular dependency audits

3. Code Review

   • All PRs reviewed before merge
   • No force-pushes to main
   • Commit signing required

4. No Telemetry

   • RedMess doesn't phone home
   • No usage tracking
   • No crash reporting
   • Audit the code yourself

# Clone and audit
git clone https://github.com/harezadmm/RedMess.git
cd RedMess

# Check for suspicious network calls
grep -r "requests.post" .
grep -r "urllib" .
grep -r "socket" .

# Review dependencies
cat requirements.txt
pip install pipdeptree
pipdeptree

# Run in isolated environment
docker build -t redmess .
docker run --rm --network none redmess

RedMess uses cryptography for:

1. API Key Storage

   • Keys encrypted at rest
   • OS keyring integration (when available)
   • Fallback: XOR obfuscation (not secure, use keyring)

2. Telegram Bot Tokens

   • Stored in config file (plaintext)
   • User responsible for file permissions
   • Recommend: chmod 600 ~/.hermes/config.yaml

3. Conversation Encryption (Optional)

   # Enable encryption
   hermes config set encryption.enabled true
   hermes config set encryption.key $(openssl rand -hex 32)

RedMess processes:

• Conversation history (locally stored)
• User IDs (Telegram, Discord)
• API keys (encrypted)

Your rights:

• Access: cat ~/.hermes/conversations/*
• Deletion: rm -rf ~/.hermes/
• Portability: Copy ~/.hermes/ directory

No data sent to maintainers unless you report a bug.

Same as GDPR. We don't collect data centrally.

RedMess may contain:

• Strong cryptography (export-controlled)
• Offensive security tools (export-controlled)

Users are responsible for compliance with:

• US Export Administration Regulations (EAR)
• International Traffic in Arms Regulations (ITAR)
• Local export control laws

Do not use RedMess in embargoed countries without proper licenses.

• Security Issues: security@redmess.dev
• General Questions: GitHub Discussions
• Urgent Matters: Telegram @harezadmm

PGP Key for security@redmess.dev:

-----BEGIN PGP PUBLIC KEY BLOCK-----
[Key would go here in real implementation]
-----END PGP PUBLIC KEY BLOCK-----

Last updated: 2026-09-04

Security is a shared responsibility. Report issues responsibly.