The Agent Skills Directory

[SAFE]: The skill provides legitimate documentation and development resources for the AutoGPT platform. The author, 'kalarislabs', is consistent with the provided vendor context and provides secure patterns for agent operation and platform management.
[EXTERNAL_DOWNLOADS]: The documentation provides instructions to clone the source code from the official AutoGPT GitHub repository ('github.com/Significant-Gravitas/AutoGPT'). This is a well-known, public repository for the project and is documented neutrally as a standard installation step.
[INDIRECT_PROMPT_INJECTION]: The platform described handles external data via webhooks and API inputs. The skill includes a detailed 'Agent Operating Procedure' that instructs the agent to validate inputs, pin parameters, and confirm identifiers before execution. These instructions serve as boundary markers to mitigate the risk of processing untrusted data.
[COMMAND_EXECUTION]: Routine commands are provided for setting up the environment using Docker Compose, installing dependencies via NPM and Poetry, and performing database migrations with Prisma. These commands are typical for the described development and production environments.
[CREDENTIALS_UNSAFE]: No hardcoded credentials or unsafe secret handling practices were found. The skill describes a secure credential management system that uses encryption and environment variables to protect provider API keys.

autogpt-agents