The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by reading configuration from .claude/dotfiles.local.md in the current working directory.
Ingestion points: Settings are read from .claude/dotfiles.local.md (project-scoped) and ~/.claude/dotfiles.local.md (global) in SKILL.md.
Boundary markers: None. The skill parses YAML frontmatter using awk and grep without explicit instructions to ignore embedded prompts in the values.
Capability inventory: The skill uses awk, grep, and chezmoi via shell execution. It returns a verified directory path that subsequent skills use for dotfiles operations (reading/writing to the system).
Sanitization: The skill validates that the resolved path exists and contains Chezmoi-specific markers (.chezmoiroot or home/ directory).

dotfiles-config