The Agent Skills Directory

[PROMPT_INJECTION]: The skill's agent definitions, such as those found in chapter-extractor.md and narrative-writer.md, contain explicit instructions for the AI to ignore content safety guidelines regarding sensitive, uncomfortable, or violent themes when performing literary analysis on fictional works.
[COMMAND_EXECUTION]: This skill functions as an installer that deploys a complex system of lifecycle hooks. These hooks execute various local shell scripts (.sh, .cmd), Python modules (.py), and Node.js scripts (.js) during session start, tool use, and compacting events. The setup process involves changing file permissions (chmod +x) and bypassing PowerShell execution policies on Windows systems.
[EXTERNAL_DOWNLOADS]: The session-start.sh hook is configured to perform automated update checks by querying the GitHub API for the latest releases from the author's repository. It also provides prompts suggesting the use of the npx package runner for tool updates.
[DATA_EXFILTRATION]: The story-researcher.md agent provides research capabilities by utilizing browser automation (agent-browser) to conduct web searches and extract textual content from external websites for use in the writing project.
[REMOTE_CODE_EXECUTION]: While primarily deploying local scripts, the skill establishes a framework where external commands are triggered automatically during the AI agent's operation, and it encourages the execution of remote code via package managers for updates.

story-setup