Skip to content

Security: OctavianTocan/ai-agent-behavior-evals

Security

SECURITY.md

Security and Privacy

This repo is designed to be public-safe.

It should not contain:

  • OpenClaw source
  • OpenClaw runtime state
  • API keys or tokens
  • Telegram chat IDs
  • private sessions
  • raw reports with verbose provider payloads
  • agent memory files

Generated reports are ignored by git. If a report is worth keeping, copy only a small sanitized excerpt into examples/ or docs/.

Before publishing changes, run:

git status --short
git diff --cached --stat
python3 -m py_compile run.py
python3 run.py --dry-run

For live confidence, also run at least one production-path smoke:

python3 run.py --case evidence-before-reassurance

There aren't any published security advisories