By pbdeuchler
Run autonomous experiment loops to optimize measurable code targets: branch git repos, generate AI hypotheses, execute bash benchmarks, log metrics with confidence scores to JSONL, commit improvements, revert failures, and resume after crashes.
npx claudepluginhub pbdeuchler/llm-plugins --plugin autoresearchSet up and run an autonomous experiment loop for any optimization target. Gathers what to optimize, then starts the loop immediately. Use when asked to "run autoresearch", "optimize X in a loop", "set up autoresearch for X", or "start experiments".
Compute and interpret MAD-based confidence scores for experiment results. Use when logging experiment results after 3+ data points to determine if improvements are real or within noise.
Git commit and revert patterns for autoresearch experiments. Use when keeping or discarding experiment results to manage git state correctly.
Parse METRIC output lines, infer units, and track primary vs secondary metrics. Use when processing experiment output from autoresearch.sh.
Manage autoresearch.jsonl logging, session initialization, segment tracking, and session recovery. Use when starting, resuming, or recording experiments.
Claude Code plugins for design, implementation, and development workflows. Largely stolen from ed3d-plugins, ToB, davebcn87 and ever so slightly modified.
Autonomous experiment loop that optimizes any measurable target. Point it at a metric and it iteratively tries ideas, benchmarks them, keeps improvements, and discards regressions -- logging everything to a structured JSONL file. Runs indefinitely or until a time/iteration limit. Each experiment executes in an isolated subagent to keep the main context clean.
/autoresearch:start [duration-minutes]
Multi-perspective engineering review of any codebase -- or a scoped subset -- from a single command. A Haiku-powered scout maps the structure, then an Opus-powered panel of staff engineers reviews sampled files across seven dimensions (correctness, consistency, simplicity, design principles, idiomatic usage, security, test quality) and returns severity-classified findings with holistic remediation prose. For large codebases, files are partitioned by module and reviewed in parallel.
/blank-slate-review:start [scope]
Tightly scoped implementation planning with a panel of specialist engineer subagents. Creates plans of 5 steps or fewer, each roughly one story point, ready to hand off to an implementer. Five specialist agents (systems performance, distributed systems, security, infra ops, product lead) evaluate approaches from their domain at specific process steps.
/quick-plan:start [basic prompt]
Executes an implementation plan end-to-end in a single session: creates a branch, implements each step with TDD, runs per-step code review (fixing all severity levels), performs a holistic final review via a multi-persona staff engineer panel (with optional dueling-model review via Codex), and opens a PR. Rejects plans too large or vague to complete in 5 steps at a high quality bar.
/one-shot:start <absolute-plan-file-path> [seed-commitish]
Opinionated development guides covering coding patterns, testing strategies, database access, and technical writing. Skills activate automatically when relevant -- functional core / imperative shell, defense in depth, property-based testing, PostgreSQL conventions, and more.
/plugin install house-style@llm-plugins
Reference skills for developer tools: ast-grep for structural code search and transformation, and qmd for searching markdown knowledge bases. Loaded automatically when relevant tool usage is detected.
/plugin install tooling@llm-plugins
/plugin marketplace add https://github.com/pbdeuchler/llm-plugins.git
All plugins are available from the llm-plugins marketplace:
/plugin install autoresearch@llm-plugins
/plugin install blank-slate-review@llm-plugins
/plugin install house-style@llm-plugins
/plugin install one-shot@llm-plugins
/plugin install quick-plan@llm-plugins
/plugin install tooling@llm-plugins
Skills and commands to perform one shot prompting in a standardized workflow
Autonomous experiment loop that optimizes any file by a measurable metric. 5 slash commands, 8 evaluators, configurable loop intervals (10min to monthly).
Autonomous experimentation skill — your AI coding agent designs experiments, tests hypotheses, discards failures, keeps wins. Runs overnight while you sleep.
Autonomous experiment loops on any codebase — one file, one metric, one loop. Based on Karpathy's autoresearch pattern.
Comprehensive UI/UX design plugin for mobile (iOS, Android, React Native) and web applications with design systems, accessibility, and modern patterns
Uses power tools
Uses Bash, Write, or Edit tools
Share bugs, ideas, or general feedback.
Qiushi Skill: methodology skills for AI agents guided by seeking truth from facts, with Claude Code, Cursor, OpenClaw, Codex, OpenCode, and Hermes guidance.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge.
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge.
Sign in to claim