The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion and processing of data from a persistent knowledge graph via Memento MCP, which presents a surface for instructions from previous sessions to influence future behavior.
Ingestion points: External data is pulled into the agent's context through tools like mcp__memento__semantic_search and mcp__memento__open_nodes as described in SKILL.md and references/memento-protocol.md.
Boundary markers: The instructions do not specify any delimiters or safety markers to differentiate between recalled data and current task instructions, nor do they advise the agent to ignore instructions embedded in memory.
Capability inventory: The skill utilizes file-writing capabilities (WORKING_STATE.md), MCP tool execution, and potential command-line operations (npx).
Sanitization: There is no evidence of sanitization, filtering, or validation of the retrieved memory content before it is processed by the LLM.
[EXTERNAL_DOWNLOADS]: The documentation encourages the user to download and install the Memento MCP server from a third-party GitHub repository (github.com/gannonh/memento-mcp). It also provides a command to install additional skills from the author's own repository using npx.
[COMMAND_EXECUTION]: The skill includes instructions to execute the npx command to add further capabilities (debugging-protocol), which involves fetching and executing code from a remote source.

memory-protocol