The Agent Skills Directory

[COMMAND_EXECUTION]: The skill orchestrates the use of CLI tools such as gh and git for repository management and leverages platform-specific tools like session_spawn and session_start_background to execute tasks within isolated VM sandboxes.
[EXTERNAL_DOWNLOADS]: The skill retrieves external content by cloning GitHub repositories using gh repo clone or git clone to perform development work within its delegated environments.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection as it processes untrusted data from external sources, including repository source code, pull request diffs, and ticket/issue descriptions.
Ingestion points: Untrusted content enters via gh repo clone, git clone, and gh pr diff operations.
Boundary markers: No specific requirements for delimiters or 'ignore' instructions are mandated for data passed into child session prompts.
Capability inventory: Delegated sessions possess full code execution, file system access, and network capabilities within their isolated sandboxes.
Sanitization: The instructions do not define sanitization or filtering procedures for data retrieved from external sources before it is processed by the agent.
[DATA_EXFILTRATION]: The skill accesses project-specific context and state stored in .kortix/memory and retrieves user/organizational information via the GitHub API to correctly route work and identify target repositories.

coding