The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of the ecc-universal package from the NPM registry as a runtime prerequisite.
[COMMAND_EXECUTION]: The skill relies on several CLI commands (ecc memory, ecc-memory-mcp, ecc-memory-vault) for its core operations, including searching, reading, and saving data to the local file system.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from a 'Memory Vault' which may contain untrusted content from previous sessions or other agents.
Ingestion points: Data is ingested through ecc memory search and ecc memory read commands (and their MCP equivalents memory_search and memory_read) as documented in SKILL.md.
Boundary markers: The instructions in SKILL.md explicitly advise the agent to treat recalled bodies as 'untrusted context' and never as 'executable instructions', providing a conceptual boundary.
Capability inventory: The skill allows for file system writes (ecc memory save, ecc memory handoff) and diagnostic checks (ecc memory doctor) as described in SKILL.md.
Sanitization: The instructions state that the runtime rejects known secret shapes, but prompt-level sanitization of retrieved markdown documents is not explicitly defined beyond the instruction to verify claims against authoritative sources.

unified-memory