The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The troubleshooting guide recommends using privileged commands like sudo systemctl start docker and sudo ufw allow 8001 to manage local system services and networking.
[COMMAND_EXECUTION]: The skill includes instructions for common administrative tasks such as terminating processes using kill -9 and managing application lifecycles via docker compose and npm scripts.
[EXTERNAL_DOWNLOADS]: The documentation directs users to clone the platform source code from the official project repository at https://github.com/Significant-Gravitas/AutoGPT.git.
[DYNAMIC_EXECUTION]: The core platform functionality relies on a graph-based execution engine that dynamically coordinates logic blocks, subgraphs, and ability scripts at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill describes an autonomous agent platform that ingests data from external sources, presenting an inherent surface for indirect prompt injection.
Ingestion points: Data enters the system context through Webhook blocks, HTTP request blocks, and external integrations with services like GitHub, Discord, and Notion.
Boundary markers: The provided code examples and documentation do not demonstrate the use of specific delimiters or system instructions to differentiate between user-supplied data and agent instructions.
Capability inventory: The platform's executor and Forge toolkit possess extensive capabilities, including network access, modular Python ability execution, and tool orchestration.
Sanitization: While referencing sanitized inputs in logging examples, the skill does not detail explicit validation or escaping mechanisms for content ingested via external blocks.

autogpt-agents