The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs users to clone source code from an external repository located at https://github.com/alash3al/stash.git. This repository is not associated with a known trusted vendor.
[REMOTE_CODE_EXECUTION]: The installation process involves downloading external code and executing it via docker compose up or go build, which facilitates the execution of remote code on the local system.
[COMMAND_EXECUTION]: Several administrative and build commands are required for setup and maintenance, including docker exec, go build, and curl for triggering consolidation pipelines or setting context.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a memory layer that ingests potentially untrusted data and re-injects it into the LLM's system prompt or context.
Ingestion points: Data enters the system via the stash_remember MCP tool and the /api/episodes HTTP endpoint as defined in SKILL.md.
Boundary markers: The instructions do not define or implement boundary markers or instructions for the agent to ignore potentially malicious content within recalled memories.
Capability inventory: The skill manages persistent storage, performs semantic searches, and interacts with external LLM providers, providing a mechanism to persist and re-trigger instructions across sessions.
Sanitization: There is no mention of sanitization, filtering, or validation mechanisms for the content stored in the memory layer before it is presented back to the AI agent.

stash-ai-memory