The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a feedback loop where the agent's own task descriptions, managed via the TodoWrite tool, are injected back into the agent's prompt context through the getAppContext callback defined in SKILL.md.
Ingestion points: The getAppContext callback in SKILL.md (Step 4) reads the current todos state and formats it into a string for the prompt.
Boundary markers: The prompt uses a simple header (Current todo list:) but lacks explicit instructions for the agent to ignore potentially malicious content within the todo items themselves.
Capability inventory: The skill allows access to Bash, Write, Edit, Read, Glob, and Grep via the allowed-tools configuration.
Sanitization: There is no sanitization or escaping applied to the item.content strings before they are joined and returned to the agent's context.
[EXTERNAL_DOWNLOADS]: The instructions in SKILL.md (Step 1) direct the agent to check for and install @tabler/icons-react using the project's package manager. This is a well-known, standard library for React icons and is consistent with the UI components provided in the skill.

integrate-todo-list