The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of a 'story' CLI tool to perform various project tasks such as 'story add', 'story links', 'story build', 'story context', and 'story reindex'. These commands are standard for managing a story-project's structure and building outputs like Twine (Twee) or Ink files.
[DYNAMIC_EXECUTION]: In environments where the 'story' CLI is not globally installed, the skill instructs the agent to execute a local JavaScript file ('story.js') using Node.js. This execution occurs from a relative path ('../story-maintenance/scripts/story.js') or a user-provided directory ('/bin/story.js'). The instructions specifically advise against using Bun to avoid the execution of untrusted 'bunfig.toml' files, demonstrating a proactive security posture.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data from story files, including 'story.md' and various chapter files in the 'chapters/' directory. This data, particularly the 'choices' frontmatter and passage prose, is used to construct the story's branching logic and context. The skill mitigates risks by scoping context ('story context') to specific story paths and providing clear instructions on handling rejoin prose.
[SAFE]: The skill follows established developer patterns for repository-based workflows. It emphasizes explicit user approval for destructive actions (like removing chapters), uses snapshots for recovery, and enforces strict rules against writing ambiguous or contradictory story state. The use of relative paths and vendor-specific scripts is consistent with the 'Story Skills' ecosystem.

interactive-fiction