agentrigs
+ submit rig
~ / rigs / vitaliikapliuk / modelharness

vitaliikapliuk/modelharness

Make every model cheaper or better. Zero-config behavioral harness for Claude Code, with a reproducible 408-run benchmark

↗ GitHub ★ 36 MIT updated 4mo ago personal setup Claude Code Claude plugin
share on X
ARCHETYPE
Pragmatist
A balanced, no-drama setup: some rules, some tools, nothing extreme.
CONTEXT TAX · EVERY TURN
~236 tokens
Featherweight · median rig: 2.3k · breakdown
GUARDRAILS
0/5
No committed guardrails · details

Copy this rig

# review before running: this installs third-party code
$ npx degit vitaliikapliuk/modelharness/agents ./rig-modelharness/agents
$ npx degit vitaliikapliuk/modelharness/commands ./rig-modelharness/commands
$ npx degit vitaliikapliuk/modelharness/skills ./rig-modelharness/skills
$ npx degit vitaliikapliuk/modelharness/hooks ./rig-modelharness/hooks

MCP servers are added to Claude Code at local scope; env vars are shown as YOUR_… placeholders — we never store values. Files are fetched with degit into a separate folder so you can review before merging.

$ npx degit vitaliikapliuk/modelharness/skills/delegation-triggers .claude/skills/delegation-triggers
$ npx degit vitaliikapliuk/modelharness/skills/memory-discipline .claude/skills/memory-discipline
$ npx degit vitaliikapliuk/modelharness/skills/verification-loop .claude/skills/verification-loop
$ curl -fsSL --create-dirs -o .claude/agents/verifier.md https://raw.githubusercontent.<redacted>.md

This rig commits no guardrails. Here is the community baseline instead — the deny/ask rules most often found across all 7,204 rigs:

{
  "permissions": {
    "deny": [
      "Read(./.env)",
      "Read(**/.env)",
      "Read(~/.ssh/**)",
      "Bash(rm -rf *)",
      "Read(**/*.pem)",
      "Bash(rm -rf /)",
      "Bash(git push --force:*)",
      "Bash(sudo *)",
      "Read(.env)",
      "Bash(rm -rf /*)",
      "Read(./.env.*)",
      "Read(~/.aws/**)",
      "Bash(git push --force*)",
      "Bash(rm -rf:*)",
      "Read(**/*.key)",
      "Read(**/.env.*)",
      "Bash(sudo:*)",
      "Bash(git reset --hard*)",
      "Bash(git reset --hard:*)",
      "Read(.env.*)"
    ],
    "ask": [
      "Bash(git push:*)",
      "Bash(git push *)",
      "Bash(git commit:*)",
      "Bash(rm *)",
      "Bash(rm:*)",
      "Bash(git rebase *)",
      "Bash(wget *)",
      "Bash(npm publish:*)",
      "Bash(git commit *)",
      "Bash(gh pr merge *)"
    ]
  }
}

Skills (3)

delegation-triggersmemory-disciplineverification-loop

Subagents (1)

verifierFresh-context verifier. Use PROACTIVELY after completing any multi-step task - verifies work against its specification by running real checks, immune to the implementer's rationalizations.

Slash commands (3)

/goal/retro/verify

Similar rigs

Remus3/Legion-Wallpaper
Self-auditing AI image restoration pipeline (super-resolution, LaMa inpainting, face repair, metric-plus-vision gate ladder), built by a multi-agent Claude Code system with independent verifiers, git-hook gates and headless background lanes
★ 1 project Claude CodeCodexGemini CLI 1 agents 16 hooks today
Automator 5.5k tok ·
leavemagic-cyber/earned-confidence
A behavior contract for AI coding agents — every rule added after a real failure. Ships with a verifier agent that cannot edit what it judges.
★ 3 project Claude Code 2 agents 2mo ago
Pragmatist 3.6k tok ·
NotRedFox/NotRedFoxs-Claude-skills
Claude Code skills: a docs fact-checker, a code auditor, a bug-memory log, a mod and more.
★ 1 project Claude Code 4 agents 4 hooks today
Automator 577 tok ·
rlaope/quoted-andrej-karpathy
An harness and behavioral guidelines to enforce reliable execution in AI coding agents.
★ 1 project Claude CodeCodexCursor 4 skills 5mo ago
Minimalist 1.5k tok ·
copied ✓