The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill features a dependency management system in deps.md that enables the fetching and execution of code from Git repositories via prose install. This is a primary function of the framework and adheres to standard development practices, including using a prose.lock file for version pinning.- [DATA_EXFILTRATION]: No unauthorized data exfiltration patterns were found. The skill includes explicit security instructions in SKILL.md and state/postgres.md to prevent the exposure of secrets, connection strings, or environment variables in logs or artifacts.- [COMMAND_EXECUTION]: The skill coordinates subagents using ProseScript choreography. These capabilities are governed by 'Shape' constraints (e.g., Shape.prohibited) that define strict behavioral boundaries for services, as detailed in guidance/authoring.md.- [PROMPT_INJECTION]: The skill processes untrusted external data, such as GitHub webhooks and incident reports, creating a surface for indirect prompt injection. However, the architecture mitigates this through subagent context isolation and the 'copy-on-return' protocol, which ensures that only declared outputs are passed between services.

open-prose