The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses scripts/generate_tech_manga.py to execute a required dependency (nanobanana.py) via subprocess.run. The command is constructed as a list, which effectively prevents shell injection. However, the search heuristic used to locate the script includes the current working directory (./nanobanana.py), which is a common but sub-optimal practice that could lead to path-hijacking if run in an untrusted directory.
[EXTERNAL_DOWNLOADS]: The documentation references and requires the installation of nanobanana-skill from a third-party GitHub repository (feiskyer/claude-code-settings). This is a legitimate dependency necessary for the skill's core image generation functionality.
[DATA_EXPOSURE]: The skill implements secure credential management by instructing users to store their Gemini API key in a local .env file with restrictive permissions (chmod 600). This is considered a safe and recommended practice for managing sensitive data.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests user-provided technical concepts to generate AI prompts and interactive HTML content. However, the risk is inherent to its primary function, and the structured nature of the planning steps provides a layer of operational predictability.

tech-manga-explainer