The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a multi-layered memory system that ingests and processes data from various local and remote sources.
Ingestion points: The agent is instructed to read context from SESSION-STATE.md, MEMORY.md, daily logs in the memory/ directory, and external stores like LanceDB and Mem0.
Boundary markers: The provided templates for memory files use Markdown headers to organize context but lack explicit instructions to the agent to treat stored content as untrusted or to ignore embedded instructions.
Capability inventory: The agent has the capability to write to the file system (via elite-memory CLI), execute shell commands (python3 memory.py), and interact with network APIs (Mem0, SuperMemory).
Sanitization: There is no evidence of automated sanitization, escaping, or validation of the content being stored in or retrieved from the memory files.
[COMMAND_EXECUTION]: The skill's functionality relies on the execution of various CLI tools and shell commands.
Execution patterns: The instructions guide the agent to use npx elite-longterm-memory, python3 memory.py, and git notes for storing and retrieving context.
Evidence: These patterns are found throughout the setup and maintenance sections of SKILL.md and README.md.
[EXTERNAL_DOWNLOADS]: The skill references external packages and services to extend its memory capabilities.
References: The documentation suggests installing the mem0ai package and using the SuperMemory API for cloud synchronization.
Source: References target the official NPM registry and established service providers, documented neutrally as intended features.
[DYNAMIC_EXECUTION]: The provided CLI utility generates executable and instructional content at runtime.
Script generation: The bin/elite-memory.js script uses fs.writeFileSync to generate SESSION-STATE.md, MEMORY.md, and daily log files based on pre-defined templates.
Evidence: This logic is contained within the init and today commands of the CLI tool.

elite-longterm-memory