The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it is designed to ingest and process untrusted external data.
Ingestion points: The mempalace mine command reads content from local project directories (~/projects/myapp) and chat logs (~/chats/).
Boundary markers: No specific delimiters or instructions to ignore embedded prompts in processed data are documented.
Capability inventory: The skill provides a suite of 19 tools for reading, writing, and searching the 'memory palace' database, as well as an MCP server for persistent access.
Sanitization: No evidence of sanitization or filtering of the ingested content is present in the provided scripts.
[COMMAND_EXECUTION]: The skill utilizes local shell commands for its core operations and automation.
The documentation instructs users to configure an MCP server using python3.11 -m mempalace.mcp_server.
The skill implements an automated 'Stop' hook in settings.json that executes hooks/mempal_save_hook.sh upon completion of agent tasks. This script runs a status check and potential context mining via the Python module.

mempalace