Zen Koan Daily

This skill generally does what it says (generate koans, an image prompt, and TTS), but there are several things to check before use: - Clarify Gemini usage: README and generate_image.py mention Google Gemini and suggest exporting GEMINI_API_KEY, but the registry metadata does not declare this credential. If you do not want to provide a Gemini key, confirm how images are generated (the shipped script only prints a prompt/command). - Cross-skill exec: generate_image.py prints a command that calls a generate_image.py under ~/clawd/skills/yumfu/, not the local script. That means image generation may rely on another skill/tool you don't have — verify or adjust the command to use a trusted local image pipeline. - Runtime bug: references/progress.json initially lacks a completed_koans array but generate_koan.py expects progress['completed_koans'] when updating — this can crash or corrupt progress on first run. Inspect and initialize progress.json or patch the script before enabling scheduled runs. - edge-tts is invoked locally (no cloud key) — ensure you trust the package you install (pipx install edge-tts) and run in an isolated environment if you are cautious. - LLM dynamic generation and optional web_search: the skill relies on the LLM to synthesize koans and optionally search the web. If you need strictly canonical or fully-accurate historical text, review generated outputs for fidelity and possible hallucinations. Recommendations: run this skill in an isolated/test environment first, review/patch the generate_image.py command and progress.json initialization, and only provide any cloud API keys (Gemini) if you understand how/where they will be used. If you want to use images but do not want to supply a Gemini key, modify the skill to call a local/approved image generator or remove the external dependency.