The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill guides the user to clone the official AutoGPT repository from GitHub (https://github.com/Significant-Gravitas/AutoGPT.git) to set up the platform backend and frontend. These downloads originate from a well-known project repository.
[COMMAND_EXECUTION]: Instructions include standard commands for building and running a local development environment, such as docker compose, npm install, and poetry run. These are necessary for the platform's installation and operation and are executed by the user during setup.
[INDIRECT_PROMPT_INJECTION]: The skill describes a system architecture designed to ingest and process external data, which is an inherent characteristic of autonomous agent platforms.
Ingestion points: Data enters the system context via webhook handlers and HTTP request blocks as described in the platform architecture in SKILL.md and advanced-usage.md.
Boundary markers: Explicit boundary markers for isolating untrusted data in LLM prompts are not detailed in these documentation snippets.
Capability inventory: The platform supports extensive capabilities including file system access, network requests, and interaction with third-party APIs through specialized blocks.
Sanitization: The documentation references the use of sanitized inputs in error tracking and logging contexts in the advanced usage guide.

agents-autogpt