Latest release: v1.0.1Download zip
Capabilities
Security Scan
OpenClaw
Suspicious
high confidencePurpose & Capability
The declared skill is only “Mirror Palace / mirror-palace,” while the included artifacts describe a full personal operating system, identity/persona, long-term memory, subagents, backups, and third-party integrations.
Instruction Scope
Root-level agent instructions attempt to control every session, require reading persona/user/memory files, and include first-run behavior for an external BOOTSTRAP.md file. This is much broader than an invocation-scoped skill.
Install Mechanism
Registry/install metadata says this is an instruction-only skill with no requirements, but the bundle includes runnable code, backup scripts, and nested skill/package directories. There is no evidence of automatic execution, but the provenance and activation boundary are unclear.
Credentials
The bundle includes personal profile, health, relationship, memory, path, and workflow context that is not proportionate to the minimal public description.
Persistence & Privilege
Artifacts describe non-negotiable memory writes, cron/heartbeat behavior, multi-agent reporting, and Telegram delivery. These are persistent/autonomous behaviors that are not disclosed by the top-level SKILL.md.
What to consider before installing
Treat this as a published personal workspace backup, not a clean scoped skill. Do not install it unless you intentionally want the Taylor/Chloe memory and operating rules in your agent context. A safe version should strip personal files, remove root AGENTS.md behavior, remove unrelated nested skills/scripts, and clearly document any memory, scheduler, cloud, or account integrations.