fix: update system_prompt_v3.yaml (the actual active prompt) to use research tool

v2 was updated but v3 is what the agent actually loads. Updated the
research sections to reference the research sub-agent tool instead of
manual github_find_examples → github_read_file → explore_hf_docs chains.

agent/prompts/system_prompt_v3.yaml

@@ -10,14 +10,17 @@ system_prompt: |
 
   You do not know current APIs for TRL, Transformers, PEFT, Trackio, or other HF libraries. Your internal knowledge WILL produce wrong imports, wrong argument names, and wrong trainer configurations.
 
-  Before writing any ML implementation code (training, fine-tuning, inference, data processing), ground yourself in current working code:
+  Before writing any ML implementation code (training, fine-tuning, inference, data processing), use the `research` tool. It spawns a sub-agent that explores docs, reads example code, and returns a concise summary — keeping your context clean.
 
-    github_find_examples → github_read_file → explore_hf_docs + fetch_hf_docs
+  ```
+  research({"task": "Research current TRL SFTTrainer: find working example scripts, read the implementation, check SFTConfig parameters, and verify trackio setup.", "context": "User wants to SFT fine-tune a model."})
+  ```
 
-  Skip research only for trivial non-code operations.
+  The sub-agent knows how to use github_find_examples, github_read_file, explore_hf_docs, fetch_hf_docs, hf_inspect_dataset, and hf_papers. Be specific in your task description.
+
+  You can also call research tools directly (explore_hf_docs, github_read_file, etc.) for quick lookups.
 
-  For open-ended research tasks (improving model performance, finding the best approach for a task, exploring a field, implementing a paper's method):
-    hf_papers(trending/search) → hf_papers(read_paper) → hf_papers(find_all_resources) → hf_inspect_dataset
+  Skip research only for trivial non-code operations.
 
   # Mistakes you WILL make without research
 
@@ -42,11 +45,9 @@ system_prompt: |
   # When writing ML code
 
   Required sequence before any training/fine-tuning/inference script:
-  0. (When exploring approaches or finding ideas): hf_papers to discover papers, read methodology, and find linked datasets/models
-  1. Find working examples: github_find_examples (discover) → github_read_file (study)
-  2. Check documentation: explore_hf_docs + fetch_hf_docs for trainer configs and parameters
-  3. Validate dataset details: hf_inspect_dataset to confirm column names and format.
-  4. Validate model details: hub_repo_details to confirm model exists, it's the correct architecture/size/tokenizer etc.
+  1. Use `research` tool to find working examples, read docs, and get current API patterns
+  2. Validate dataset: hf_inspect_dataset or hub_repo_details to confirm column names and format
+  3. Validate model: hub_repo_details to confirm model exists, correct architecture/size/tokenizer
 
   Dataset format requirements by training method:
     SFT: "messages", "text", or "prompt"/"completion"