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 using hosted LLM (Cerebras) — requires CEREBRAS_API_KEY from https://cloud.cerebras.ai

Parallel exploration of implementation approaches - run multiple variants simultaneously and let tests determine the winner

Iterative artifact refinement - hone any artifact or workspace over multiple rounds using criteria-driven judge feedback, runnable evaluators, and focused directional improvements

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

Automatically updates terminal title with emoji + project + topic context

Agentic binary reverse engineering for ELF binaries on ARM64, ARMv7, x86_64 - structured analysis with radare2, Ghidra, GDB, and dynamic tracing

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

Prepares codebases for public/open-source release and audits them for openness with full lifecycle prep or standalone review

PR health assistant - monitors CI, triages review comments, fixes issues with prevention. Triggers on 'PR', 'CI', 'checks', 'review comments', 'workflow failed'.

Parallel exploration framework that uses diverse perspective panels for both generating approaches and evaluating implementations, then synthesizes the best insights from all variants into the winner.

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

Tiered-delegation task execution - a planner model (Sonnet) turns a spec into a contract that pins every cross-cutting decision plus self-contained sprints, a fast cheap model (Haiku) executes the bulk in parallel and self-verifies against the gate, and a scoped fix runs only on failure. Benchmarked ~64% cheaper than Opus building the same spec, at equal gate quality. General-purpose: code and non-code tasks.