Claude Code Skills

Two skills.
Zero friction.
Infinite experiments.

Drop a trigger phrase into Claude Code. AutoResearch sets up your project, generates program.md, runs the baseline — and starts iterating while you sleep.

Get autoresearch-setup → autoresearch-skill-optimizer ↓
2
Skillsincluded in this release
5
Questionsautoresearch-setup asks
100
MetricSQS target for skills
0
Config filesneeded to install
1
Commandto launch overnight run
SKILL 01 / 02

autoresearch-setup

One-call AutoResearch onboarding for any ML project. Answer 5 questions — Claude handles everything: git branch, program.md, baseline run, and a copy-paste agent prompt you fire before bed.

ML Projects HuggingFace PyTorch Any Framework
/autoresearch-setup
click to copy trigger Install Skill →
What It Does
  • Asks 5 targeted questions about your training script, GPU, metric, research direction, and agent preference — all in one message
  • Detects your setup: pure AutoResearch repo, custom train.py, conda/pip project, or Apple Silicon (MLX)
  • Creates a git branch autoresearch/[date] and initializes results.tsv
  • Generates a customized program.md from framework-specific templates
  • Runs the baseline experiment to establish your starting metric
  • Returns an exact copy-paste agent prompt for Claude Code, API, or Codex
Trigger Phrases — Type Any
"set up autoresearch" COPY
"add autoresearch to my project" COPY
"autoresearch setup" COPY
"set up autoresearch for my LoRA project" COPY
/autoresearch-setup COPY
What Claude Does — 4 Phases
0
Intake — 5 Questions in One Message
Asks PROJECT (script path), GPU (hardware), METRIC (val_loss, accuracy), DIRECTION (what to explore), AGENT (Claude Code / API / Codex). Waits for all answers before continuing.
1
Mode Detection
Mode A (AutoResearch repo) → Mode B (custom train.py, uv-compatible) → Mode C (pip/conda, wrapper needed) → Mode D (Apple Silicon, redirects to autoresearch-mlx).
2
Setup Execution
Creates git branch, writes program.md from template, initializes results.tsv header, adapts metric parsing, runs baseline experiment, captures starting metric value.
3
Agent Prompt Generation
Produces exact copy-paste text for your chosen agent (Claude Code / API / Codex) with project-specific context baked in. Shows morning-checklist commands before signing off.

Skills in /Users/meetdeshani/Desktop/skills/ are auto-loaded by Claude Code on every session — no extra config needed. The skill is already available if you're on Meet's machine.

Verify installed
# In any Claude Code session, just type:
autoresearch setup

# Or with a hint:
set up autoresearch for my LoRA fine-tuning project

# Skill file location (already exists):
/Users/meetdeshani/Desktop/skills/autoresearch-setup/SKILL.md

Copy or symlink the skill folder into any project so it's available when Claude Code runs in that directory.

bash
# Option A — hard copy (static snapshot)
cp -r /Users/meetdeshani/Desktop/skills/autoresearch-setup/ \
  /path/to/your/project/skills/autoresearch-setup/

# Option B — symlink (stays in sync with source)
mkdir -p /path/to/your/project/skills
ln -s /Users/meetdeshani/Desktop/skills/autoresearch-setup/ \
  /path/to/your/project/skills/autoresearch-setup

# Then open Claude Code in the project dir — skill is live

Deploy to a VPS running OpenClaw or any remote Claude agent.

bash
# Upload to OpenClaw skills directory on VPS
scp -r /Users/meetdeshani/Desktop/skills/autoresearch-setup/ \
  root@your-vps:/opt/services/openclaw/data/agents/main/agent/skills/

# Restart OpenClaw to pick up new skill
ssh root@your-vps "docker restart openclaw"

# For any Claude API agent — paste SKILL.md into system prompt
cat /Users/meetdeshani/Desktop/skills/autoresearch-setup/SKILL.md
SKILL 02 / 02

autoresearch-skill-optimizer

Applies Karpathy's AutoResearch loop to SKILL.md files. Instead of training a neural network, it improves prompt quality — one surgical edit at a time, keeping gains and reverting regressions, using SQS (Skill Quality Score) as the metric.

SKILL.md Files Self-Improving AI Prompt Engineering
/autoresearch-skill-optimizer
click to copy trigger Install Skill →
What It Does
  • Scores every SKILL.md in your skills repo against the SQS 5-axis rubric (T, P, E, R, C — 20 pts each, target ≥ 80)
  • Runs up to 5 hypothesis-driven experiments per skill: proposes one change, applies it, re-scores, keeps if SQS improves, reverts if it doesn't
  • Logs every experiment to .autoresearch/results.tsv with hypothesis, axis, expected gain, and actual delta
  • Stops early when a skill reaches SQS ≥ 80 — "research complete"
  • Can target one skill (optimize higgsfield-prompt) or all at once (optimize my skills)
  • Produces a session report with before/after table, delta, experiments kept vs reverted
Trigger Phrases — Type Any
"optimize my skills" COPY
"autoresearch-skill-optimizer report" COPY
"show me SQS scores for all skills" COPY
"optimize higgsfield-prompt" COPY
/autoresearch-skill-optimizer COPY
The SQS Metric — Skill Quality Score (0–100)
T
Trigger Clarity
20 pts — Would Claude fire at exactly the right moment?
P
Protocol
20 pts — Can Claude follow steps mechanically without guessing?
E
Examples
20 pts — Real input → output examples present?
R
Rules
20 pts — Clear DO/DON'T constraints with edge cases?
C
Conciseness
20 pts — Zero filler — every word earns its place?
Skill Before → After Delta
meet-documents72 → 81+9
higgsfield-prompt65 → 74+9
MeetSkillDesktopIndex78 → 83+5

Already installed on Meet's machine. Open any Claude Code session and use a trigger phrase.

Verify installed
# In any Claude Code session:
optimize my skills

# For a specific skill only:
autoresearch-skill-optimizer meet-documents

# Report only — no changes:
autoresearch-skill-optimizer report

# Skill file location:
/Users/meetdeshani/Desktop/skills/autoresearch-skill-optimizer/SKILL.md

Copy or symlink to use in any project with its own skills directory.

bash
# Hard copy
cp -r /Users/meetdeshani/Desktop/skills/autoresearch-skill-optimizer/ \
  /path/to/your/project/skills/autoresearch-skill-optimizer/

# Or symlink (stays in sync)
mkdir -p /path/to/your/project/skills
ln -s /Users/meetdeshani/Desktop/skills/autoresearch-skill-optimizer/ \
  /path/to/your/project/skills/autoresearch-skill-optimizer

Deploy to VPS running OpenClaw or paste SKILL.md into any Claude API system prompt.

bash
# Upload skill to OpenClaw on VPS
scp -r /Users/meetdeshani/Desktop/skills/autoresearch-skill-optimizer/ \
  root@your-vps:/opt/services/openclaw/data/agents/main/agent/skills/

# Restart to pick up
ssh root@your-vps "docker restart openclaw"

# For Claude API — read SKILL.md and paste into system prompt
cat /Users/meetdeshani/Desktop/skills/autoresearch-skill-optimizer/SKILL.md

Comparison

Which Skill Do You Need?

They work independently but compose well — set up AutoResearch on a project, then optimize the skills that power it.

Question autoresearch-setup autoresearch-skill-optimizer
I want to run AutoResearch on my ML project ✓ This one
I want to improve my Claude Code SKILL.md files ✓ This one
Works on HuggingFace Trainer, PyTorch, Axolotl ✓ All frameworks
Works on Apple Silicon (no NVIDIA GPU) ✓ Mode D → MLX ✓ No GPU needed
Creates git branch + baseline run
Runs hypothesis → edit → score → keep/revert loop ✓ (on model code) ✓ (on SKILL.md)
Produces a copy-paste overnight agent prompt
Logs results to TSV file ✓ results.tsv ✓ .autoresearch/results.tsv

Quick Start

From zero to overnight run
in four steps

Works with any Claude Code session. No config files, no environment variables, no YAML.

STEP 01
Install the skills

Skills are in /Desktop/skills/ — auto-loaded by Claude Code. No extra setup if you're on Meet's machine. Otherwise copy the SKILL.md folder into your project or Claude Code skills path.

Check if loaded
# Open Claude Code and type:
what skills do you have?
STEP 02
Onboard your project

Navigate to your ML project in Claude Code, then trigger the setup skill. Claude will ask 5 questions and do the rest.

In Claude Code
set up autoresearch on my project

# Claude will ask:
1. What is your training script?
2. What hardware? (H100 / A100 / 4090 / M2)
3. What is your validation metric?
4. What should the agent explore?
5. Which AI agent? (Claude Code / API / Codex)
STEP 03
Copy the agent prompt

After setup, Claude prints an exact agent launch prompt. Copy it. Paste it into Claude Code (or your chosen agent) and hit Enter.

Example output
═══════════════════════════════════
  COPY THIS INTO CLAUDE CODE
═══════════════════════════════════
Read program.md and train.py.
Run: uv run python train.py
Parse val_loss from output.
If improvement → git commit "exp-N"
Else → git checkout train.py
Repeat 100x. Log to results.tsv.
═══════════════════════════════════
STEP 04
Sleep. Check in the morning.

The agent runs 100 experiments overnight. In the morning, check what improved.

Morning checklist
# See what was kept
grep "KEPT" results.tsv

# Review all changes
git diff main..autoresearch/$(date +%b%d | tr A-Z a-z)

# Update program.md with learnings
update program.md based on results.tsv
Ready
Open Claude Code.
Type one phrase.

Both skills are ready. The loop starts tonight.

Copy "set up autoresearch" → Read the guide ↗ See all applications ↗
Copied to clipboard