Popular repositories Loading

1. loopkit loopkit Public

   33 battle-tested skills + minimal .claude harness for any coding agent (Claude Code, Cursor, Codex, Gemini CLI).

   Shell 732 129

2. adversarial-contract-gate adversarial-contract-gate Public

   Self-evaluation is a trap — grade against a contract with a separate evaluator. Built from Anthropic's 'Build Agents That Run for Hours' (39:00). No deps, ./run.sh, 9 tests.

   Python 24 3

3. loop-graph-harness loop-graph-harness Public

   Working agent pipeline built from timestamped Anthropic lectures

   Python 7 3

4. graph-vs-vector-provenance graph-vs-vector-provenance Public

   On a multi-hop question a bag-of-words vector baseline returns a similar-but-wrong chunk; a toy knowledge graph walks the answer and prints the path. Built from Zach Blumenfeld's Intro to GraphRAG.…

   Python 7 2

5. memory memory Public

   The memory system Anthropic engineers describe publicly, packaged as a drop-in .claude/. MEMORY.md + consolidator + overnight dreamer subagent + session hooks.

   Shell 4 5

6. file-memory-vs-vector file-memory-vs-vector Public

   When a fact changes between sessions, a vector store returns the similar-but-stale version; file memory overwrites one address and reads the current truth with file:line provenance. Built from 'Tur…

   Python 2 1