AgentCharter is a framework project with no runtime code. If you discover a security risk in its rule design or templates (e.g., privilege escalation vectors, data leak scenarios), please:

• Do not open a public Issue
• Contact the project maintainer (via GitHub private message if no email is published)
• We will acknowledge within 48 hours and begin remediation

AgentCharter's risk surface is primarily in:

• Git isolation rules: ensuring non-TPM Agents cannot execute git commands
• Permission table accuracy: incorrect read/write directions in ACTIONS.md could allow Agents to accidentally modify others' files
• Context injection safety: context/ memory files containing sensitive credentials must be sanitized by the TPM before injection

These risks are mitigated by the user project's TPM through CHARTER.md. The framework provides only the rule templates.