agentrigs
+ submit rig
~ / rigs / tigerless-labs / autoharness

tigerless-labs/autoharness

Autoharness — a self-learning skill layer for Claude Code — distills skills from your real sessions, updates them as you work, and prunes the ones that stop getting used. No daemon, no benchmark.

↗ GitHub ★ 7,647 mit updated 7d 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
~2.6k tokens
Moderate · median rig: 2.2k · breakdown
GUARDRAILS
0/5
No committed guardrails · details

Copy this rig

# review before running: this installs third-party code
$ npx degit tigerless-labs/autoharness/agents ./rig-autoharness/agents
$ npx degit tigerless-labs/autoharness/skills ./rig-autoharness/skills
$ npx degit tigerless-labs/autoharness/hooks ./rig-autoharness/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 tigerless-labs/autoharness/skills/learn .claude/skills/learn
$ curl -fsSL --create-dirs -o .claude/agents/curator.md https://raw.githubusercontent.com/tigerless-labs/autoharness/main/agents/curator.md
$ curl -fsSL --create-dirs -o .claude/agents/reflector.md https://raw.githubusercontent.com/tigerless-labs/autoharness/main/agents/reflector.md

This rig commits no guardrails. Here is the community baseline instead — the deny/ask rules most often found across all 6,413 rigs:

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

MCP servers (1)

serversourceest. tokens
stage_skill
env: PYTHONPATH
 local / custom 2.5k

Skills (1)

learn

Subagents (2)

curator
model: haiku
Periodic consolidation pass — fold narrow agent-created skills into class-level umbrellas. Proposes intents only, never writes to disk.
reflector
model: haiku
Distill a finished episode into skill changes aligned with the existing library. Compare-first preference, generation stays open; proposes intents only, never writes to disk.

Similar rigs

nagisanzenin/engram
Evidence-based learning engine — first-principles curricula, free-recall verification with receipts, FSRS-scheduled memory, and explorable artifacts. Learn anything; keep it.
★ 1,439 project Claude Code 3 skills 1mo ago
Minimalist 185 tok ·
kirilxd/claude-tutor
Turn Claude Code into your personal tutor — personalized learning plans, adaptive quizzes, SM-2 spaced repetition, and a web dashboard. Works with any topic.
★ 135 project Claude Code 5 skills 3mo ago
Pragmatist 1.5k tok ·
john-wilmes/claude-agentic-coding-playbook
Evidence-based practices for LLM-assisted development. Install scripts, Claude Code configuration, and a research-backed best practices guide with 34 verified citations.
★ 2 project Claude Code 6 skills 6mo ago
Pragmatist 3.9k tok ·
devduck123/swe-os
Engineering judgment for people and their coding agents. One set of Markdown files: a learning site for humans, instructions for agents.
★ 0 project Claude CodeCodex 7 skills today
Pragmatist 1.2k tok ·
copied ✓