The Agent Skills Directory

[SAFE]: The skill provides documentation, workflow templates, and reference materials for managing Git worktrees.
[INDIRECT_PROMPT_INJECTION]: The skill establishes a pattern where the agent reads metadata from .ai-context.json and README.working-tree.md within the user's workspace to calibrate its behavior (e.g., switching between 'main', 'feature', 'bugfix', 'experiment', and 'review' modes).
Ingestion points: Workspace files .ai-context.json and README.working-tree.md (referenced in SKILL.md and references/best-practices.md).
Boundary markers: No explicit prompt boundaries or sanitization instructions for the metadata content are provided.
Capability inventory: The skill is designed to work with Git management tools (/working-tree:*) and provides guidance for development workflows.
Sanitization: The skill does not provide instructions for sanitizing or validating the contents of the ingested metadata files before the agent processes them.

worktree-guide