[!WARNING] Load order is critical. This skill MUST be loaded after anibon-timestamper-local. Loading this skill first on a local model (Gemma, Qwen) will cause an immediate abort due to prefill budget exhaustion from the combined skill text size. Always load anibon-timestamper-local first, then load this skill.

🔄 Anibon Timestamper Handoff Protocol

When processing long video transcripts (>2 hours) on local models (Ollama/Gemma/Qwen), the conversation context window can overflow after processing several sequential chunks, causing the model to become slow, unresponsive, or forgetful.

Use this handoff protocol to save the session state, flush the context, and resume processing in a brand-new conversation session.

🗺️ Plugin Directory Map (Do NOT use ls)

You already know where everything is. Resolve [SKILL_ROOT] and [WORKSPACE] from the state file or from Step 0 of anibon-timestamper-local.

• Main Scripts: [SKILL_ROOT]/scripts/prepare_video.py
• DB + Check Scripts: [SKILL_ROOT]/scripts/ (fetch_fgo_db.py, fetch_ygo_db.py, check_sections.py)
• Workspace: [WORKSPACE] — stored in state file as workspace_path
  • [WORKSPACE]/chunks/chunk_XX.txt
  • [WORKSPACE]/chunk_outputs/chunk_XX_output.md
  • [WORKSPACE]/anibon_timestamper_state.json