The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on environment variables ($MCP_CALL and $MEMPALACE_MCP_SCRIPT) to execute shell commands. This pattern allows the agent to run external Python scripts, which is a significant capability that depends on the integrity of the local environment setup.
[PROMPT_INJECTION]: The skill creates an indirect prompt injection surface by storing and retrieving arbitrary text in a persistent database.
Ingestion points: Data is ingested into the persistent storage via the mempalace_add_drawer, mempalace_kg_add, and mempalace_diary_write tools.
Boundary markers: The skill does not provide specific instructions or delimiters to isolate retrieved memories, increasing the risk that stored malicious instructions could be followed by the agent when they are fetched later via mempalace_search or mempalace_diary_read.
Capability inventory: The skill possesses extensive capabilities, including command execution and filesystem access (via the referenced scripts), which could be abused if an injection is successful.
Sanitization: There is no evidence of input validation or output sanitization to prevent the storage or execution of malicious payloads embedded in the memory data.

mempalace