The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill includes shell scripts (scripts/generate_game.sh, scripts/generate_playground.sh) designed to dynamically generate HTML and JavaScript artifacts that serve as interactive reading playgrounds.\n- [COMMAND_EXECUTION]: The utility scripts perform standard file system operations, including directory creation (mkdir -p) and writing multi-line code blocks to local files using heredocs.\n- [INDIRECT_PROMPT_INJECTION]: The utility scripts accept user input for output filenames via the read command without performing validation or path sanitization. This creates a vulnerability surface for path traversal attacks.\n
Ingestion points: The OUTPUT_FILE variable in both generation scripts, populated by user/agent input.\n
Boundary markers: Absent; the input flows directly into the file path construction.\n
Capability inventory: Capability to write arbitrary files (cat > "$OUTPUT_PATH") and create directories within the execution environment.\n
Sanitization: Absent; the scripts do not strip path separators or validate that the output remains within the intended directory, potentially allowing the agent to overwrite sensitive files if executed with malicious parameters.

reading-teacher