The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local project-level commands such as lint, typecheck, and grep to verify code changes and perform repository searches during task implementation.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests and processes task descriptions and documentation from external sources. Ingestion points: Task content is retrieved via the Linear get_issue tool or read from local Markdown files in the docs/tasks/ directory. Boundary markers: Safety is reinforced through a Goal Articulation Gate and a mandatory Pre-Submission Checklist to validate task intent and outcomes. Capability inventory: The skill is authorized to modify files (Edit), update task management metadata (update_issue), and execute shell commands. Sanitization: It enforces a human-in-the-loop requirement for operations classified as critical and prohibits automated code commits, ensuring manual oversight of all modifications.

ln-401-task-executor