The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill relies on external Node.js packages agentdb and agentic-flow that are not from pre-approved trusted sources. These packages are downloaded from the NPM registry during initialization and usage.
[COMMAND_EXECUTION]: The instructions utilize npx agentdb@latest to execute code directly from the NPM registry. This includes initialization, data migration, and the setup of an MCP server (npx agentdb@latest mcp), which allows external code to run within the agent's environment.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to store and retrieve "experiences" and "trajectories" in a vector database to influence future agent decisions, which creates a potential injection surface.
Ingestion points: Untrusted data is stored via rb.insertPattern and retrieved via rb.retrieveWithReasoning as shown in SKILL.md.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard potential commands embedded within the stored memory data.
Capability inventory: The skill performs file writes to local database files (.agentdb/reasoningbank.db) and executes shell commands via npx.
Sanitization: The code snippets do not show any sanitization, validation, or filtering of the content being stored in or synthesized from the database.

ReasoningBank with AgentDB