2389-research

CSS development workflows with Tailwind composition, semantic naming, and dark mode by default

Firebase project workflows including setup, features, debugging, and validation

Token-efficient code generation pipeline - parallel implementation with hosted LLM (Cerebras) for ~60% token savings. Includes MCP server.

Parallel exploration of implementation approaches - implements multiple variants simultaneously and lets tests determine the winner

Iterative artifact refinement with investigation-first judge board - constructs problem-specific judges that read the code, understand the problem, and propose evidence-based improvements

Decision-making through deliberation - seeking unity through discernment rather than consensus through debate

Automatically updates terminal title with emoji + project + topic context for quick visual cues when switching terminals

Agentic binary reverse engineering for ELF binaries on ARM64, ARMv7, x86_64 - hypothesis-driven analysis with radare2, Ghidra, GDB, QEMU

Dispatch panels of specialized subagents to review projects — expert audits, first-impression personas, task-completion flows, and pedantic nitpicks

Prepare codebases for public/open-source release and audit them for openness - full lifecycle prep or standalone review

PR health assistant - monitors CI, triages review comments, fixes issues with systematic prevention. Uses gh CLI and gh-pr-review extension.

Parallel exploration powered by diverse perspectives - independent agent panels propose, build variants in worktrees, review, pick a winner, and synthesize the best of all variants into the final result

Take digital drugs that modify Claude's behavior with prompt injection. Effects activate immediately and persist across sessions.

MCP server for agent social media platform - enables AI agents to interact in team-based discussions

A comprehensive MCP server that provides Claude with private journaling, semantic search, resources, and guided prompts for reflection and learning

Tiered-delegation execution where Sonnet plans a spec into sprints, a cheap Haiku agent builds and self-verifies against the gate, and a scoped Sonnet fix runs only on failure - benchmarked ~64% cheaper than Opus at equal quality.