The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external task trackers and user-provided ideas to populate decision graphs and task records.\n
Ingestion points: Remote tracker records (labeled graphify:map), project configuration files like AGENTS.md, and user-supplied descriptions in SKILL.md.\n
Boundary markers: The skill utilizes Markdown structural elements (headings, lists) to organize content, though it lacks explicit "ignore embedded instructions" delimiters for data retrieved from remote sources.\n
Capability inventory: The skill can execute shell commands, run project task runners (pnpm, python, bun), and perform file system and git operations across multiple scripts (e.g., references/local-graph.md, references/research.md).\n
Sanitization: No specific sanitization logic is described for data processed from external trackers.\n- [DYNAMIC_EXECUTION]: The skill generates and executes dynamic code as part of its prototyping workflow to answer design and logic questions.\n
It creates self-contained HTML/JS logic prototypes (references/logic.md) and UI variations (references/ui.md) that are executed in the user's environment.\n
It uses project-specific task runners (e.g., pnpm, python, bun) to launch these experimental artifacts.\n- [COMMAND_EXECUTION]: The skill employs system-level commands to interact with the host environment and present artifacts to the user.\n
It uses tools like open (macOS) or xdg-open (Linux) to display generated HTML visualizations and prototypes (references/html.md).

graphify