Agent Claude Code
Reviews code for architectural patterns, abstraction levels, and reuse of existing code. Use to catch over-engineering, missed abstractions, or pattern violations.
6 7mo ago A 35 tokens
original MIT
Agent Claude Code
Groups code changes by architectural concern for replay guide generation. Takes diff analysis output and returns logical clusters.
6 7mo ago A 26 tokens
original MIT
Agent Claude Code
Evaluates current implementation against a design spec or requirements doc. Use periodically during vibe coding to detect drift, missing features, or invariant violations. Returns a focused report, not code.
6 7mo ago A 40 tokens
original MIT
Agent Claude Code
Analyzes git diffs to extract structural changes for replay guide generation. Returns classified changes with dependency mappings.
6 7mo ago A 25 tokens
original MIT
Agent Claude Code
Runs build checks and tests, reporting results clearly. Use to catch regressions without polluting main context with verbose test output.
6 7mo ago A 28 tokens
original MIT
Agent Claude Code
Generates a Human Replay Guide from a vibe coding session. Use after completing exploratory work in a sandbox to create an optimal, ordered guide for manual integration. Invoke with the session name and base commit.
6 7mo ago A 45 tokens
original MIT
Agent Claude Code
Research worker for vibe-coding sessions. Explores codebase, understands patterns, and gathers context before implementation begins. Returns structured findings to the orchestrator.
6 7mo ago A 37 tokens
original MIT
step-writer
08Agent Claude Code
Writes detailed replay steps for a cluster of changes. Focuses on teaching the human, not just showing code.
6 7mo ago A 26 tokens
original MIT
Agent Claude Code
Implementation worker for vibe-coding sessions. Receives specific phase objectives and implements them autonomously. Returns a concise summary to the orchestrator.
6 7mo ago A 34 tokens
original MIT
vibe-coding
10Skill Claude CodeCodex
Enables autonomous exploration mode. Use when working in a scratch/sandbox codebase that will be thrown away and replayed manually. Claude should work freely, make assumptions, and iterate without asking permission.
6 7mo ago A 45 tokens
original MIT
vibe-prep
11Skill Claude CodeCodex
How to prepare for a vibe coding session that facilitates human replay.
6 7mo ago A 18 tokens
original MIT
MCP server Claude CodeCodexCursor +2
A graph database and analysis tool for AI assistants via MCP. Runs locally from the mcp-graph-engine Python package.
6 7mo ago A
tokens not measured
original MIT