The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the configuration of background tasks that execute commands. It provides specific instructions to verify command existence and scope permissions narrowly (e.g., gh pr list instead of *).
[PERSISTENCE]: While the skill creates persistent scheduled tasks, it utilizes the platform's official sidecar mechanism and requires the user to manually enable the task through an official dashboard, preventing silent persistence.
[PRIVILEGE_ESCALATION]: The skill proactively warns against granting excessive privileges like root file access (/) or wildcard command access (*). It instructs the agent to define the minimum necessary permissions based on a test run.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided text to create a task prompt. While this creates a surface where a user could define a malicious task, the skill itself provides instructions for creating safe, self-contained prompts and relies on the agent's internal safety filters during the generation and execution phases.

automation