The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a persistent memory system that ingests, stores, and later retrieves arbitrary text data, creating an attack surface for indirect prompt injection.
Ingestion Points: Untrusted data enters the system through the content parameter in mempalace_add_drawer, the object parameter in mempalace_kg_add, and the entry parameter in mempalace_diary_write (all in SKILL.md).
Boundary Markers: There are no defined delimiters or instructions to treat stored content as data rather than instructions when retrieved.
Capability Inventory: The skill is capable of executing shell commands using the python3 $MCP_CALL pattern across all 19 tools defined in SKILL.md.
Sanitization: The skill lacks explicit sanitization or filtering of stored content before it is re-injected into the agent's context via tools like mempalace_search or mempalace_kg_query.
[COMMAND_EXECUTION]: The skill operates by executing local Python scripts through shell commands (python3 $MCP_CALL "python3 -u $MEMPALACE_MCP_SCRIPT" ...). This involves the execution of external scripts referenced by environment variables defined in the skill's metadata.

mempalace