Agent Console is a local operator surface for AI agent runtimes. It can sit near profiles, chat history, cron jobs, skills and provider credentials, so public packaging must be conservative.

The free repository is intended for local development and self-hosted evaluation. If you expose it beyond localhost, add your own authentication, TLS, network allowlisting and process supervision.

Agent Console should preserve these defaults:

• Do not return raw secrets to the browser.
• Do not commit .env files, private profile files, SQLite runtime state, logs, certificates or keys.
• Keep sessions, cron jobs and skills read-only unless a future change has explicit review.
• Keep delegated mission execution opt-in. Mission execution is disabled by default and should only be enabled on a private network after reviewing local CLI and MCP permissions.
• If missions are enabled, keep dangerous shell, file-write and MCP tools constrained. Do not rely on bypassPermissions alone when MCP servers are configured.
• Do not publish private Perseon agent identities, persona canon, internal memory, private prompts, customer data or host-specific runtime data.
• Redact screenshots before they reach GitHub, docs, issues or social previews.

Please do not open a public issue with sensitive details.

Email the maintainer or contact Perseon Labs through the official Perseon AI contact path with:

• a short description of the issue
• reproduction steps
• affected endpoint or file
• impact assessment
• whether any secret, token, private profile or customer data may have been exposed

We will acknowledge valid reports as quickly as possible and coordinate a fix before public disclosure.

Before making this repo public, verify:

• git status --short has no accidental runtime files.
• git ls-files does not include certs, keys, logs, state databases or private profile files.
• Screenshots in screenshots/ use example data only.
• .env.example contains placeholders only.
• README copy does not expose private agents, internal operating canon or host-specific secrets.
• python3 -m py_compile agent_console.py passes.
• Local /health and /agent-config return safe metadata only.