The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied topics and interview answers, persisting them into local markdown files such as plan.md and session-log.md. The agent is instructed to load these files to resume sessions, creating a surface where malicious input provided during discovery could influence agent instructions in subsequent interactions.\n
Ingestion points: Discovery Interview responses and topic slugs in SKILL.md.\n
Boundary markers: No explicit delimiters or instructions are provided to ignore potentially malicious embedded instructions when re-loading session logs.\n
Capability inventory: File system read/write access and shell command execution (open, xdg-open).\n
Sanitization: The skill lacks instructions to sanitize user input before persistence or re-interpolation into the prompt context.\n- [COMMAND_EXECUTION]: The skill directs the agent to execute shell commands (open on macOS and xdg-open on Linux) to open generated HTML lesson files. This involves invoking external utilities based on agent-generated file paths.\n
Evidence: Found in resources/html-lessons.md instructions for opening the file for the learner.\n- [DYNAMIC_EXECUTION]: The skill generates self-contained HTML files containing embedded JavaScript for interactive quiz elements. The generation of scripted content that the user is then prompted to execute (via a browser) is a form of dynamic code generation.\n
Evidence: Specified in resources/html-lessons.md for 'reveal-on-click' quiz logic.\n- [DATA_EXPOSURE]: The skill accesses specific directories within the user's home folder (~/.claude/learning/ and ~/.claude/skills/) to manage session state and discover curriculum materials. While intended for state management, access to the ~/.claude/ directory is a potential exposure point for agent configuration.\n
Evidence: Logic for checking existing progress and matching skills in SKILL.md.

teach-me