The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple shell commands to facilitate installation and configuration, including uv tool install, pip install, mempalace init, and chmod +x to make hook scripts executable.
[DATA_EXFILTRATION]: The skill targets sensitive local data by mining conversation transcripts from ~/.claude/projects/. While the documentation emphasizes a local-first architecture with no cloud dependency, the instruction to read and process these transcripts represents a significant data access capability.
[EXTERNAL_DOWNLOADS]: The skill downloads external software, specifically the mempalace package and a ~300MB embedding model, from remote repositories and registries during the setup process.
[COMMAND_EXECUTION]: The instructions require modifying .claude/settings.local.json to register shell hooks (mempal-stop-hook.sh and mempal-precompact-hook.sh) that automatically execute commands during specific agent lifecycle events.
[COMMAND_EXECUTION]: The skill uses the GitHub CLI (gh api) to fetch repository content and decodes it from Base64 directly into the terminal, which is an unusual method for viewing documentation.

arise-mempalace