By toqitahamid
Autonomous ML research agent for Claude Code: HF ecosystem tools, GitHub ML-search, and the ml-intern workflow discipline. Runs training on your HPC.
npx claudepluginhub toqitahamid/ml-intern-plugin --plugin ml-internAutonomous ML research agent for Claude Code. Adds a skill that crawls ML literature, audits HF datasets, writes pre-flight-validated training scripts, and submits to your HPC. Based on the ml-intern agent by 🤗 Hugging Face.
uv: curl -LsSf https://astral.sh/uv/install.sh | sh/plugin marketplace add toqitahamid/ml-intern-plugin
/plugin install ml-intern@ml-intern-plugin
/reload-plugins
The first /mcp check may time out while uv creates the Python venv
(~10–30s). Run /mcp again and it will connect.
Add to ~/.zshrc or ~/.bashrc, then restart Claude Code:
export HF_TOKEN="hf_..." # https://huggingface.co/settings/tokens
export GITHUB_TOKEN="ghp_..." # https://github.com/settings/tokens (public_repo)
Without tokens, HF tools rate-limit on public data and fail on gated content; GitHub tools rate-limit after a handful of calls.
Dataset inspection:
/ml-intern find a small image-classification dataset on Hugging Face and show me its schema
Research workflow:
/ml-intern plan a SFT recipe for a 3B model on a reasoning benchmark — start from the literature
HPC script generation:
/ml-intern write a SLURM sbatch script for a 7B SFT run on 4x A100 with Trackio logging
The skill auto-triggers on plain ML prompts too; the slash command is the explicit escape hatch.
ml-intern skill — research-first workflow, tuned for HPC
submission via SLURM or PBS. No HF Jobs, no remote sandbox.explore_hf_docs, fetch_hf_docs,
find_hf_api, hf_inspect_dataset, hf_repo_files, hf_repo_git,
hf_papers, github_find_examples, github_list_repos,
github_read_file./ml-intern — force-activates the skill for a turn./ml-intern-doctor — diagnoses install and env issues.Run /ml-intern-doctor — it prints a ✓/✗ table for uv, Python, venv,
tokens, and MCP server health.
/mcp shows ml_intern_tools failed. On first install, wait 30
seconds and retry. If it still fails, uv is either missing (run
command -v uv) or not on Claude Code's PATH — symlink it into
~/.local/bin.
401 / rate-limit errors. Tokens missing. Set them in your shell profile and restart Claude Code.
Broke after a plugin update. Stale venv:
rm -rf ~/.claude/plugins/cache/ml-intern-plugin/ml-intern/*/.venv,
then /mcp.
git clone https://github.com/toqitahamid/ml-intern-plugin
cd ml-intern-plugin
uv sync --extra dev
uv run pytest
Issues and PRs welcome at
github.com/toqitahamid/ml-intern-plugin.
Highest-value contributions: Windows support (a scripts/run-mcp.cmd
sibling of the POSIX launcher), bug reports, and skill-prompt
refinements.
Team-oriented workflow plugin with role agents, 27 specialist agents, ECC-inspired commands, layered rules, and hooks skeleton.
Admin access level
Server config contains admin-level keywords
External network access
Connects to servers outside your machine
Requires secrets
Needs API keys or credentials to function
Upstash Context7 MCP server for up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your LLM context.
Complete collection of battle-tested Claude Code configs from an Anthropic hackathon winner - agents, skills, hooks, rules, and legacy command shims evolved over 10+ months of intensive daily use
Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques
Tools to maintain and improve CLAUDE.md files - audit quality, capture session learnings, and keep project memory current.
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
Requires secrets
Needs API keys or credentials to function