Skip to content

Orchestra-Research/AI-Research-SKILLs / 0-autoresearch-skill

autoresearch

Verified

Orchestrates end-to-end autonomous AI research projects using a two-loop architecture. The inner loop runs rapid experiment iterations with clear optimization targets. The outer loop synthesizes results, identifies patterns, and steers research direction. Routes to domain-specific skills for execution, supports continuous agent operation via Claude Code /loop and OpenClaw heartbeat, and produces research presentations and papers. Use when starting a research project, running autonomous experiments, or managing a multi-hypothesis research effort.

templates/research-state.yaml

View on GitHub
# Research State — Central Project Tracking
# Copy this template to your project root and fill in as you go.
# Updated by the agent after each experiment and reflection.

project:
  title: ""
  question: ""                    # The core research question
  status: active                  # active | paused | concluded
  started: ""                     # ISO date
  domain: ""                      # e.g., "mechanistic interpretability", "RL training"

literature:
  key_papers: []
  # - id: "liu2025superposition"
  #   title: "Superposition Yields Robust Neural Scaling"
  #   authors: "Liu et al."
  #   year: 2025
  #   relevance: "Proves ETF structure in LM heads"
  open_problems: []               # Gaps identified from literature
  evidence_gaps: []               # What's missing in the field

hypotheses:
  # List of all hypotheses, active and completed
  # - id: H1
  #   statement: "Testable claim with clear prediction"
  #   status: pending             # pending | active | supported | refuted | inconclusive
  #   motivation: "Why this is worth testing"
  #   parent: null                # null for root, parent ID (e.g., H1) for sub-hypotheses
  #   priority: medium            # high | medium | low

experiments:
  proxy_metric: ""                # What we're optimizing and how to compute it
  baseline_value: null            # Starting point
  best_value: null                # Best achieved so far
  total_runs: 0
  trajectory: []
  # - run_id: "run_001"
  #   hypothesis: "H1"
  #   metric_value: null
  #   delta: null                 # Change from baseline
  #   wall_time_min: null
  #   change_summary: ""
  #   timestamp: ""

outer_loop:
  cycle: 0                       # How many outer loop reflections so far
  last_direction: null            # deepen | broaden | pivot | conclude
  last_reflection: ""             # Brief summary of last reflection decision

workspace:
  # Track key resource locations
  findings: "findings.md"
  log: "research-log.md"
  literature_dir: "literature/"
  experiments_dir: "experiments/"
  to_human_dir: "to_human/"
  paper_dir: "paper/"
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 323kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 323kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 323kUpdated Claude CodeCodexVerifiedInstall