The Agent Skills Directory

Task Logic and Reliability: The skill implements a sophisticated linting system (Phase 3) designed to identify and correct patterns that cause AI agents to stall or fail during autonomous tasks, such as ordering constraints or wall-clock dependencies. This logic is intended to increase the reliability of agent sessions.
File System Access: The skill generates markdown files in a specific local subdirectory (.amplifier/goals/). This allows for persistent storage of goal definitions while keeping them within a controlled project scope for easier management and versioning.
Human-in-the-Loop Design: The workflow explicitly requires the user to review the generated file before recommending the use of the /goal command. By prompting the user to "run it now" or "adjust anything first," the skill ensures user oversight and prevents the automatic execution of unverified logic.
Input Processing: The skill ingests conversation transcripts and user arguments to extract context. While these are external inputs, the skill converts them into descriptive instructions rather than executable code, significantly reducing the surface area for injection attacks.

goalify