Caveat verdict
multi-user-privacy
Multi-user privacy isolation system for agent memory files with per-user session management, quota enforcement, and admin routing; all file/process access is for the legitimate stated purpose of isolating user data within the agent platform.
⚠ Flagged for review — coarse, uncorroborated signal, not a confirmed exploit. Review the config yourself before installing.
Automated static analysis — not a human review. Caveat flags capabilities, not confirmed intent, and can produce false positives. Disagree with this verdict? Use Dispute below.
Permission integrity
agent_memory
Findings (9)
Recursive delete from root or home — destructive command
docs/VIP-AGENT-SETUP.md · code · rm -rf ~
Pipe to python — executes piped content as Python code
docs/PROJECT_SUMMARY.md · prose · downgraded · | Python
References sudo — requests elevated privileges
monitoring/QUICKSTART.md · code · sudo
Possible hardcoded credential
scripts/quota-manager.js · prose · downgraded · token=')) {
options.tokenQuota = parseInt(arg.split(
References agent memory files
SKILL.md · code · MEMORY.md
References child_process — can spawn system processes
auto-mount.js · prose · downgraded · child_process
Sets world-executable permissions
UPGRADE_v0.4.0.md · code · chmod 750
Opens WebSocket connection
docs/PROJECT_SUMMARY.md · prose · downgraded · WebSocket
Node http/https module — low-level network access
monitoring/monitor.js · prose · downgraded · require('https')
Why the tier is capped
Execution sink present in raw bytes (Hard Floor: class B/D). Final tier capped at Caution — cannot be lifted by any downgrade, example-payload opt-in, or allowlist.
Permissions & capabilities
No declared permissions — minimal attack surface.
agent_memoryfile_readprocess_exec Is this flag fair?
Thanks — recorded.