feat: add claw-code benchmarking project to GTD

gtd.org

@@ -62,6 +62,19 @@ Slim down the org-agent microkernel by moving non-essential cognitive functions
 *** TODO Extract Vector Embedding Algorithms to a Skill (embedding.lisp)
 *** TODO Extract Sparse Tree Context Pruning Strategies to a Skill (context.lisp)
 
+** NEXT Project: Cognitive Benchmarking (Claw-Code Distillation)
+:PROPERTIES:
+:ID:       proj-claw-code-benchmarking
+:PROJECT_PATH: $PROJECTS_DIR/benchmarking
+:PSF-STATE: A: DEMAND
+:END:
+Systematically study the `claw-code` (Rust) and leaked Claude Code architectures to improve the `org-agent` kernel and skills.
+
+*** TODO Map Claw-Code Tool-Use Protocol (How it handles multi-tool steps)
+*** TODO Distill Claw-Code File Editing Techniques (Surgical diffing vs overwriting)
+*** TODO Analyze Claw-Code History Pruning (How it stays within context limits)
+*** TODO Write a comprehensive Note in notes/ on "Lessons from the Rust Agent Explosion"
+
 ** Agora: Decentralized Social Network
 :PROPERTIES:
 :ID:       agora-project