If you discover a security vulnerability in Chitin Shell, please report it responsibly.

DO NOT open a public GitHub issue for security vulnerabilities.

Email: security@chitin.id

Include:

• Description of the vulnerability
• Steps to reproduce
• Potential impact
• Suggested fix (if any)

• Acknowledgment: Within 48 hours
• Initial assessment: Within 5 business days
• Fix timeline: Depends on severity (Critical: 7 days, High: 14 days, Medium: 30 days)

• Bypassing process isolation (accessing credentials from the LLM container)
• Policy engine bypass (executing actions without proper verification)
• Output sanitization bypass (secrets leaking through to the LLM)
• Signature forgery or verification bypass
• Rate limiter bypass
• Audit log tampering

• The LLM producing malicious Intents (this is expected — the Policy Engine catches them)
• Denial of service on the Policy Engine (availability, not security)
• Issues in dependencies (report to the dependency maintainer)
• Social engineering attacks on human approval flow

Chitin Shell's security model is documented in ARCHITECTURE.md. Key principles:

1. Zero-Knowledge Agent: The LLM never has access to credentials
2. Intent, Not Action: The LLM produces structured requests, not raw API calls
3. Defense in Depth: Multiple independent security layers
4. Fail-Closed: Unknown actions default to Tier 3 (human approval required)

We thank the following researchers for responsibly disclosing vulnerabilities:

(None yet — be the first!)