redis-security/examples/README.md

Disposable Redis command fixture

MIT, Copyright (c) 2026 BB Skills. Review both Python scripts before use.

This experiment exercises native Redis commands and selected boundaries suggested by the official skills. It does not run an AI client or certify a complete skill, TLS, production firewall, persistence, replication, Redis Cluster, JSON, Vector Set, Java/Jedis or redis-py behavior.

Run on a disposable Linux Docker host that you control. The runner refuses existing fixture names, creates an internal Docker network, publishes no ports, uses tmpfs for server data, and removes only resources it creates. Generated passwords are passed through memory and standard input; no real account or credentials are required. Container-only 0.0.0.0 with protected mode and authentication on this isolated network is not a production bind/firewall template.

The recorded server and Python image identities are in native-redis-evidence.json. The original lab used an already-cached BB Skills application image solely as a Python runtime. For independent reproduction, acquire and review a Python image with the same Python version, then replace the runner's two bb-skills:20261004-v41 references with that local runtime image. Keep the network, tmpfs and cleanup safeguards. Review the server image version and license separately: MIT here covers the skill sources and BB Skills fixture, not the Redis server binary. Server/runtime binaries are not redistributed.

Execute python3 run-redis-fixture.py only after reviewing these prerequisites. Reproduction generates a new time, image identity and source hash if you adapt the runtime selection. The evidence records TTL windows rather than an exact clock-dependent second count. It contains synthetic key names and values only. SCAN returns names; prefix ACLs control data access rather than necessarily hiding names. ACL SETUSER changes are additive unless relevant rules are reset, and marking a user off does not close all already-authenticated connections.