Last Updated: October 28, 2025 Version: 2.0.0

DO NOT open public issues for security vulnerabilities.

Instead:

1. Email: Create a GitHub Security Advisory (preferred)

   • Go to: Repository → Security → Advisories → New draft security advisory
   • Or email: [Your security contact email]

2. Include:

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

3. Response Time: We aim to respond within 48 hours

Secrets Are Never Committed:

• ✅ All credentials use GitHub Secrets vault
• ✅ No hardcoded API keys or tokens
• ✅ .env files are gitignored
• ✅ Regular security audits performed

Secrets Used in Workflows:

• GITHUB_TOKEN - Auto-provided by GitHub (minimal permissions)
• CLAUDE_CODE_OAUTH_TOKEN - Claude Code authentication (scoped)
• PROJECTS_TOKEN - Project board access (scoped: repo + project only)

Audit Status: ✅ Last audited October 28, 2025 - No exposed credentials (See Audit)

For health-sdk-builder and healthcare applications:

Compliance Frameworks:

• ✅ HIPAA (US) - Privacy Rule, Security Rule, Breach Notification
• ✅ GDPR (EU) - Articles 5-11, Data Subject Rights
• ✅ DSGVO (Germany) - German-specific requirements
• ✅ PTV 10 (Germany) - Psychotherapy standards

Data Protection:

• ✅ Encryption at rest (AES-256)
• ✅ Encryption in transit (TLS 1.2+)
• ✅ Audit logging (all PHI/PII access)
• ✅ Access controls (role-based)
• ✅ Data minimization
• ✅ Retention policies
• ✅ Right to erasure

Clinical Safety:

• ✅ Crisis detection (suicide, self-harm, harm to others)
• ✅ Emergency escalation workflows
• ✅ Mandatory reporting compliance
• ✅ Professional boundaries
• ✅ Evidence-based practices only

Documentation: See HEALTHCARE_COMPLIANCE_GUIDE.md

4-Layer Security Model:

Layer 1: GitHub Permissions

# Only team members can trigger workflows
if: contains(fromJSON('["OWNER", "MEMBER", "COLLABORATOR"]'),
             github.event.comment.author_association)

Layer 2: Tool Restrictions

# Allowlist specific commands only
claude_args: '--allowed-tools "Bash(gh issue:*),Bash(gh pr:*)"'
# Blocks: git push, rm -rf, curl, wget, etc.

Layer 3: Token Scoping

• GITHUB_TOKEN: Minimal permissions (read by default)
• PROJECTS_TOKEN: repo + project only (no admin)
• CLAUDE_CODE_OAUTH_TOKEN: Claude operations only

Layer 4: Branch Protection

• Required pull request workflow
• Status checks required (claude-review)
• No force pushes allowed
• Admin enforcement
• Conversation resolution required

Kill Switch (.github/WORKFLOW_KILLSWITCH):

• Master OFF switch for all workflows
• Current status: DISABLED (from Oct 24 incident)
• Instant shutdown capability (no code changes needed)

Emergency Cleanup (.github/EMERGENCY_CLEANUP.sh):

• Bulk issue closure for spam cleanup
• Dry-run mode for safety
• Documented procedures

Emergency Procedures: See .github/GITHUB_WORKFLOWS_GUIDE.md

Never Commit:

• ❌ API keys (sk-ant-*, any provider)
• ❌ GitHub tokens (ghp_, github_pat_)
• ❌ OAuth tokens
• ❌ Passwords or credentials
• ❌ .env files with secrets
• ❌ Private keys (.key, .pem files)
• ❌ Patient data (PHI/PII)

Always Use:

• ✅ .env.example templates (no real values)
• ✅ GitHub Secrets for CI/CD
• ✅ .gitignore for sensitive files
• ✅ Placeholder values in examples

Patient Data Protection:

• ✅ Never commit patient data
• ✅ Use anonymized examples only
• ✅ Implement encryption (AES-256 minimum)
• ✅ Audit all data access
• ✅ Comply with HIPAA/GDPR/DSGVO

Generated Applications:

• ✅ Include compliance by default
• ✅ Document security requirements
• ✅ Provide encryption examples
• ✅ Include audit logging
• ✅ Add crisis detection (mental health apps)

Safe Skill Creation:

• ✅ No secrets in SKILL.md files
• ✅ No hardcoded credentials in Python files
• ✅ Use environment variables for configuration
• ✅ Document security requirements
• ✅ Include security best practices in generated code

Testing:

• ✅ Use test API keys (rotated after testing)
• ✅ Never commit test credentials
• ✅ Use separate testing repository
• ✅ Clean up after testing

Current Versions Receiving Security Updates:

Upgrade Recommendation: Always use latest version (2.0.x)

What: Skills can execute Python code (Code Execution Tool required) Risk: Malicious skills could execute harmful code Mitigation:

• ✅ Only use trusted skills (from this repository or verified sources)
• ✅ Review skill code before importing
• ✅ Code Execution runs in sandboxed environment (Anthropic-managed)
• ✅ No filesystem access outside sandbox

Recommendation: Review SKILL.md and Python files before importing unknown skills

What: GitHub workflows can modify issues, PRs, project boards Risk: Malicious workflow changes could spam or delete content Mitigation:

• ✅ Kill switch for emergency shutdown
• ✅ Branch protection (workflow changes require PR + review)
• ✅ Tool allowlists (restrict dangerous commands)
• ✅ Permission scoping (minimal required)
• ✅ Team-only access (@claude mentions)

Recommendation: Never merge workflow changes without review

What: Generated apps handle sensitive patient data (PHI/PII) Risk: Data breach, compliance violations, patient harm Mitigation:

• ✅ HIPAA/GDPR/DSGVO compliance built-in
• ✅ Encryption templates provided
• ✅ Audit logging examples included
• ✅ Crisis detection for mental health apps
• ✅ Clear disclaimers (not replacement for professional care)

Recommendation:

• Clinical validation required before patient use
• Legal review for compliance
• Professional oversight mandatory

• Review all workflow files for understanding
• Configure GitHub Secrets (never hardcode)
• Enable branch protection on main
• Review SECURITY_AUDIT.md
• Understand kill switch mechanism

• Review SKILL.md content
• Check Python files (if any)
• Verify source is trusted
• Check for hardcoded credentials
• Test in isolated environment first

• Complete HIPAA/GDPR/DSGVO compliance review
• Clinical validation by healthcare professionals
• Legal review for regulatory compliance
• Security audit (encryption, access controls, audit logs)
• Penetration testing (if handling real patient data)
• Privacy policy and terms of service
• Incident response plan
• Data breach notification procedures

• Security Audit: .github/SECURITY_AUDIT.md
• Workflows Guide: .github/GITHUB_WORKFLOWS_GUIDE.md
• Healthcare Compliance: HEALTHCARE_COMPLIANCE_GUIDE.md

• GitHub Security: https://docs.github.com/en/code-security
• HIPAA: https://www.hhs.gov/hipaa/
• GDPR: https://gdpr-info.eu/
• Claude Security: https://www.anthropic.com/security

Security Issues: Use GitHub Security Advisory (preferred) General Questions: Open issue with question label Healthcare Compliance: Document in issue with documentation label

Repository Security:

• ✅ No exposed credentials (audited October 28, 2025)
• ✅ Proper secrets management
• ✅ Minimal permissions enforced
• ✅ Emergency controls in place

Healthcare Applications:

• ✅ HIPAA compliance templates
• ✅ GDPR/DSGVO compliance frameworks
• ✅ Clinical safety protocols
• ✅ Evidence-based practices

Code Quality:

• ✅ Type hints (Python)
• ✅ Input validation
• ✅ Error handling
• ✅ Security best practices

Weekly:

• Monitor failed workflow runs
• Review GitHub Actions logs
• Check for unusual activity

Monthly:

• Review secrets expiration
• Audit new workflow changes
• Update dependencies

Quarterly:

• Rotate PROJECTS_TOKEN
• Comprehensive security audit
• Review permissions and access
• Update security documentation

Annually:

• Full security assessment
• Penetration testing (if applicable)
• Compliance review (HIPAA/GDPR)
• Update security policy

If Security Issue Discovered:

1. Immediate: Activate kill switch (disable workflows)
2. Assess: Determine scope and impact
3. Contain: Revoke compromised credentials
4. Remediate: Fix vulnerability
5. Communicate: Notify affected users (if applicable)
6. Document: Post-incident report
7. Prevent: Update procedures to prevent recurrence

Emergency Contacts: See .github/GITHUB_WORKFLOWS_GUIDE.md

For Healthcare Applications:

• HIPAA (US): 45 CFR Parts 160, 162, 164
• GDPR (EU): Regulation (EU) 2016/679
• DSGVO (Germany): Bundesdatenschutzgesetz (BDSG)
• PTV 10 (Germany): Psychotherapeutenvereinbarung

Responsibility: Users are responsible for ensuring compliance in their deployments. This repository provides templates and guidance, not legal advice.

This repository is secure and safe to:

• ✅ Use in production
• ✅ Share publicly
• ✅ Accept community contributions
• ✅ Deploy in enterprise environments

With proper:

• ✅ Secrets configuration (GitHub Secrets)
• ✅ Access controls (team permissions)
• ✅ Regular audits (quarterly minimum)
• ✅ Healthcare validation (for medical apps)

Last Audited: October 28, 2025 Next Audit: January 28, 2026 Status: ✅ Secure

For questions or concerns, please use GitHub Security Advisory or open an issue.