The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill is designed to fetch and install executable content from a remote community repository (open-space.cloud) using the openspace-download-skill command.
[REMOTE_CODE_EXECUTION]: It implements an autonomous "grounding agent" via the execute_task tool, which can retrieve, interpret, and apply logic from remote sources to perform multi-step operations.
[COMMAND_EXECUTION]: The skill promotes the use of CLI tools (openspace, openspace-download-skill) that execute bash commands directly, bypassing some agent protocol overhead but increasing the risk of unmonitored shell execution.
[DATA_EXFILTRATION]: The upload_skill functionality allows the agent to send local directory contents to a cloud community server, which could be exploited to exfiltrate sensitive files or configuration if the agent is directed to an incorrect path.
[PROMPT_INJECTION]: The system is vulnerable to indirect prompt injection because it searches and ingests community-contributed skills that may contain malicious instructions designed to hijack the agent's behavior during task execution.
[DYNAMIC_EXECUTION]: The "self-evolving" nature of the skill involves the fix_skill and execute_task mechanisms, which perform runtime modifications to instructions and scripts, creating a surface for unpredictable code behavior.

openspace