{
 "schema_version": 1,
 "generated": "2026-10-19",
 "repo": "google-research/harness2",
 "name": "harness2",
 "owner": "google-research",
 "slug": "google-research-harness2",
 "github": "https://github.com/google-research/harness2",
 "what": "A framework that lets an AI agent keep improving its own setup (prompts, tools, memory, skills) while the model itself stays unchanged.",
 "why": "about +853 stars a day since launch, 11k total, launched 13 days ago, talked about in 3 AI sources we track",
 "language": "Python",
 "requirements": [
  "Linux, Python 3.12 and uv for installation",
  "No credentials needed for the local demo (harness2-demo)",
  "Model-backed runs need Node.js 20+, the Google Cloud CLI, a Vertex AI project and model IDs in .env, and gcloud application-default login",
  "LAB needs Pandoc and Poppler; WorkBuddy-Bench needs Docker",
  "Model-backed benchmark runs incur API costs"
 ],
 "audience_note": "For builders who run LLM agents on professional tasks and want to improve the prompts, tools, memory and skills around a fixed model without fine-tuning it.",
 "risk": "dual-use",
 "section": "weekly",
 "weekly_rank": 2,
 "category": "Software & DevOps",
 "url": "https://dedupedai.com/use-cases/google-research-harness2/",
 "use_case_count": 8,
 "use_cases": [
  {
   "id": "google-research-harness2--audit-which-harness-edits-actually-stuck",
   "title": "Audit which harness edits were adopted and why",
   "best": true,
   "url": "https://dedupedai.com/use-cases/google-research-harness2/audit-which-harness-edits-actually-stuck/",
   "json": "https://dedupedai.com/use-cases/u/google-research-harness2/audit-which-harness-edits-actually-stuck.json"
  },
  {
   "id": "google-research-harness2--try-self-improving-agent-demo-no-keys",
   "title": "Watch an agent improve its own setup in 5 minutes, no API key",
   "best": false,
   "url": "https://dedupedai.com/use-cases/google-research-harness2/try-self-improving-agent-demo-no-keys/",
   "json": "https://dedupedai.com/use-cases/u/google-research-harness2/try-self-improving-agent-demo-no-keys.json"
  },
  {
   "id": "google-research-harness2--improve-office-web-code-agent-workbuddy",
   "title": "Boost an office, web or coding agent on WorkBuddy tasks",
   "best": false,
   "url": "https://dedupedai.com/use-cases/google-research-harness2/improve-office-web-code-agent-workbuddy/",
   "json": "https://dedupedai.com/use-cases/u/google-research-harness2/improve-office-web-code-agent-workbuddy.json"
  },
  {
   "id": "google-research-harness2--test-harness-refinement-on-legal-tasks",
   "title": "Test whether harness edits lift your agent on legal tasks",
   "best": false,
   "url": "https://dedupedai.com/use-cases/google-research-harness2/test-harness-refinement-on-legal-tasks/",
   "json": "https://dedupedai.com/use-cases/u/google-research-harness2/test-harness-refinement-on-legal-tasks.json"
  },
  {
   "id": "google-research-harness2--benchmark-agent-harness-without-answer-leaks",
   "title": "Run a clean benchmark with network limits against rubric leaks",
   "best": false,
   "url": "https://dedupedai.com/use-cases/google-research-harness2/benchmark-agent-harness-without-answer-leaks/",
   "json": "https://dedupedai.com/use-cases/u/google-research-harness2/benchmark-agent-harness-without-answer-leaks.json"
  },
  {
   "id": "google-research-harness2--compare-parallel-vs-sequential-recursion",
   "title": "Compare parallel and sequential refinement on the same task",
   "best": false,
   "url": "https://dedupedai.com/use-cases/google-research-harness2/compare-parallel-vs-sequential-recursion/",
   "json": "https://dedupedai.com/use-cases/u/google-research-harness2/compare-parallel-vs-sequential-recursion.json"
  },
  {
   "id": "google-research-harness2--swap-codex-executor-into-agent-loop",
   "title": "Run harness refinement with the Codex executor",
   "best": false,
   "url": "https://dedupedai.com/use-cases/google-research-harness2/swap-codex-executor-into-agent-loop/",
   "json": "https://dedupedai.com/use-cases/u/google-research-harness2/swap-codex-executor-into-agent-loop.json"
  },
  {
   "id": "google-research-harness2--resume-interrupted-multi-task-refinement-run",
   "title": "Resume a long refinement run after it stops midway",
   "best": false,
   "url": "https://dedupedai.com/use-cases/google-research-harness2/resume-interrupted-multi-task-refinement-run/",
   "json": "https://dedupedai.com/use-cases/u/google-research-harness2/resume-interrupted-multi-task-refinement-run.json"
  }
 ],
 "disclaimer": "Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk."
}