### Install Site Dependencies

Source: https://github.com/hughyau/academicforge/blob/site-first/CONTRIBUTING.md

Run this command to install necessary dependencies for the site.

```bash
npm run site:install
```

--------------------------------

### Install OpenFold3

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/openfold3/SKILL.md

Install the OpenFold3 package with cuequivariance support. Ensure you have Python 3.10+ and CUDA 12.1+ installed.

```bash
pip install 'openfold3[cuequivariance]==0.4.1'
```

--------------------------------

### Install Evo 2

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/evo2/SKILL.md

Install the evo2 package using pip. Model weights are automatically downloaded from Hugging Face on first use.

```bash
pip install evo2
# Weights pulled from Hugging Face on first model load.
```

--------------------------------

### Local Installer Smoke Tests (Bash)

Source: https://github.com/hughyau/academicforge/blob/site-first/README_en.md

Execute the local installer smoke tests using the Bash script.

```bash
"D:\\Application\\Git\\bin\\bash.exe" scripts/tests/forge-install-local-registry.sh
```

--------------------------------

### Verify Installed Skills on Linux/macOS

Source: https://github.com/hughyau/academicforge/blob/site-first/site/public/agents.md

After installation, use this command to list the installed skills in the respective tool's skill directory. Adjust the path based on the tool used.

```bash
ls .claude/skills/   # or .opencode/skills/ or .codex/skills/
```

--------------------------------

### Install AcademicForge on macOS/Linux

Source: https://github.com/hughyau/academicforge/blob/site-first/README_en.md

Use this command to download and run the installer script for AcademicForge on macOS or Linux systems. Specify the desired tool and skills.

```bash
cd your-project
curl -sSL https://raw.githubusercontent.com/HughYau/AcademicForge/refs/heads/site-first/scripts/forge-install.sh | bash -s -- \
  --tool claude \
  --skills humanizer,superpowers
```

--------------------------------

### Test Installer Behavior with Bash

Source: https://github.com/hughyau/academicforge/blob/site-first/CONTRIBUTING.md

If installer behavior is modified, run this bash script to test local registry installation.

```bash
"D:\Application\Git\bin\bash.exe" scripts/tests/forge-install-local-registry.sh
```

--------------------------------

### Install AcademicForge on Windows PowerShell

Source: https://github.com/hughyau/academicforge/blob/site-first/README_en.md

This script downloads the PowerShell installer for AcademicForge, then executes it to install the specified tool and skills. The temporary script file is removed afterwards.

```powershell
cd your-project
$script = Join-Path $PWD 'forge-install.ps1'
Invoke-WebRequest -Uri 'https://raw.githubusercontent.com/HughYau/AcademicForge/refs/heads/site-first/scripts/forge-install.ps1' -OutFile $script
& $script -Tool claude -Skills 'humanizer,superpowers'
Remove-Item $script
```

--------------------------------

### Compute Environment Details Example

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/compute-env-setup/SKILL.md

Example of how to record details for a specific compute environment, including its resolution method, resource tier, weight storage, GPU range, validation status, and potential issues.

```markdown
### env: proteomics-gpu
how: conda env "proteomics-gpu" on host          # on Slurm: apptainer $SCRATCH/images/proteomics-gpu.sif
                                                 # on a bridge runner: <runner-path>, image <ref>
tier: {cpus: 8, mem_gib: 64, gpus: 1}            # on Slurm also: partition, account, time, gres string
weights: CHAI_DOWNLOADS_DIR=/scratch/weights/chai (12 GB; purge-window 30d)
sm_range: sm_80..sm_90
validated: <date> (kernel-witness + agent-follows-doc clean)
gotcha: <any diagnosing-failures row hit on THIS provider>
```

--------------------------------

### Install Claude Science Skills

Source: https://github.com/hughyau/academicforge/blob/site-first/README_en.md

Installs the 'claude-science' skill collection using a curl command. This command fetches and executes an installation script.

```bash
curl -sSL https://raw.githubusercontent.com/HughYau/AcademicForge/refs/heads/site-first/scripts/forge-install.sh | bash -s -- --tool claude --skills claude-science
```

--------------------------------

### Example Commit Message Format

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/skill-creator/SKILL.md

Illustrates a format for example commit messages, differentiating between input and output for clarity.

```markdown
## Commit message format

**Example 1:**
Input: Added user authentication with JWT tokens
Output: feat(auth): implement JWT-based authentication
```

--------------------------------

### Test Installer Behavior with PowerShell

Source: https://github.com/hughyau/academicforge/blob/site-first/CONTRIBUTING.md

If installer behavior is modified, run this PowerShell script to test local registry installation.

```powershell
pwsh -File scripts/tests/forge-install-local-registry.ps1
```

--------------------------------

### Install and Run ProteinMPNN

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/proteinmpnn/SKILL.md

Installs necessary packages, clones the ProteinMPNN repository, and runs the protein_mpnn_run.py script with specified parameters. Ensure you are in the cloned repository's directory before running.

```bash
pip install torch numpy   # if not already present
git clone --depth 1 https://github.com/dauparas/ProteinMPNN.git proteinmpnn
cd proteinmpnn
python protein_mpnn_run.py \
  --pdb_path backbone.pdb --pdb_path_chains "A" \
  --out_folder out --num_seq_per_target 16 --sampling_temp "0.1"
```

--------------------------------

### Linux/macOS Install Command

Source: https://github.com/hughyau/academicforge/blob/site-first/site/public/agents.md

Use this command on Linux or macOS to install selected AI skills. Ensure you replace placeholders with actual tool and skill IDs.

```bash
curl -sSL https://raw.githubusercontent.com/HughYau/AcademicForge/refs/heads/site-first/scripts/forge-install.sh | bash -s -- \
  --tool <claude|opencode|codex> \
  --skills <id1,id2,...>
```

--------------------------------

### Install Dependencies and Run LigandMPNN

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/ligandmpnn/SKILL.md

Installs required Python packages and clones the LigandMPNN repository. It also includes a sed command to fix a numpy compatibility issue and then executes the main run script with specified model parameters, input files, and output configurations.

```bash
pip install torch numpy biopython ProDy ml_collections dm-tree
git clone --depth 1 https://github.com/dauparas/LigandMPNN.git ligandmpnn
cd ligandmpnn
sed -i 's/np.int\b/np.int64/g' openfold/np/residue_constants.py   # repo pins numpy 1.23; alias removed in >=1.24
bash get_model_params.sh ./model_params
python run.py \
  --model_type ligand_mpnn \
  --checkpoint_ligand_mpnn ./model_params/ligandmpnn_v_32_010_25.pt \
  --pdb_path complex.pdb \
  --out_folder out \
  --batch_size 8 --number_of_batches 4 \
  --temperature 0.1 \
  --fixed_residues "A45 A46 A47 A48"
```

--------------------------------

### Install RDKit Dependencies

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/remote-compute-modal/SKILL.md

Install necessary libraries for RDKit to function correctly, especially when using `rdkit.Chem.Draw`. The `chemistry_gpu` environment includes these by default.

```bash
apt_install("libxrender1", "libxext6")
```

--------------------------------

### Verify AcademicForge Installation

Source: https://github.com/hughyau/academicforge/blob/site-first/README_en.md

Check the respective skill directories to confirm that AcademicForge skills have been successfully installed for Claude, OpenCode, or Codex.

```bash
ls .claude/skills/
```

```bash
ls .opencode/skills/
```

```bash
ls .codex/skills/
```

--------------------------------

### Install Dependencies and Clone ProteinMPNN

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/solublempnn/SKILL.md

Before running SolubleMPNN, ensure PyTorch and NumPy are installed. Then, clone the ProteinMPNN repository to access the necessary scripts and models. The `cd` command is crucial as the runner uses relative imports.

```bash
pip install torch numpy   # if not already present
git clone --depth 1 https://github.com/dauparas/ProteinMPNN.git proteinmpnn
cd proteinmpnn
```

--------------------------------

### Standard p5.js Canvas Setup

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/algorithmic-art/SKILL.md

Standard p5.js structure for setting up the canvas and implementing the generative algorithm. The draw function can be static or animated.

```javascript
function setup() {
  createCanvas(1200, 1200);
  // Initialize your system
}

function draw() {
  // Your generative algorithm
  // Can be static (noLoop) or animated
}
```

--------------------------------

### Windows PowerShell Install Command

Source: https://github.com/hughyau/academicforge/blob/site-first/site/public/agents.md

Use this command on Windows PowerShell to install selected AI skills. It downloads the script, executes it, and then removes the script file. Replace placeholders with actual tool and skill IDs.

```powershell
$script = Join-Path $PWD 'forge-install.ps1'
Invoke-WebRequest -Uri 'https://raw.githubusercontent.com/HughYau/AcademicForge/refs/heads/site-first/scripts/forge-install.ps1' -OutFile $script
& $script -Tool <claude|opencode|codex> -Skills '<id1,id2,...>'
Remove-Item $script
```

--------------------------------

### Feedback JSON Structure

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/skill-creator/SKILL.md

Example structure of the feedback.json file, containing user reviews for different runs.

```json
{
  "reviews": [
    {
      "run_id": "eval-0-with_skill",
      "feedback": "the chart is missing axis labels",
      "timestamp": "..."
    },
    { "run_id": "eval-1-with_skill", "feedback": "", "timestamp": "..." },
    {
      "run_id": "eval-2-with_skill",
      "feedback": "perfect, love this",
      "timestamp": "..."
    }
  ],
  "status": "complete"
}
```

--------------------------------

### Apply Okabe-Ito Palette to Plotly

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/scientific-visualization/references/color_palettes.md

Example of how to use the Okabe-Ito palette for discrete color scales in Plotly. Ensure Plotly is installed.

```python
import plotly.graph_objects as go

okabe_ito_plotly = ['#E69F00', '#56B4E9', '#009E73', '#F0E442',
                     '#0072B2', '#D55E00', '#CC79A7']
fig = go.Figure()
# Apply to discrete color scale
```

--------------------------------

### Initialize New Artifact Project

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/web-artifacts-builder/SKILL.md

Run this script to set up a new React project for your artifact. It configures the project with necessary frontend technologies and component libraries.

```bash
bash scripts/init-artifact.sh <project-name>
cd <project-name>
```

--------------------------------

### Get Compute Provider Configuration

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/remote-compute-modal/SKILL.md

Call this function to retrieve the current compute provider configuration, including environment, app name, and egress mode. If it raises an error, the configuration has changed since the kernel started, and the kernel must be restarted.

```python
# ── compute_provider tool ──
compute_provider_config()
# → {'provider': 'byoc:modal', 'environment': 'team-bio',
#    'app_name': 'claude-science-default', 'egress_mode': 'allowlist',
#    'config_hash': '93e1a4c20b7f'}
```

--------------------------------

### List and Build Modal Environments

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/remote-compute-modal/SKILL.md

Utilize the `compute_provider` tool to list available environments and build a specific environment with hydration. This helps in finding suitable environments and preparing them for use.

```python
# compute_provider tool, provider='modal'
need = {'transformers', 'torch'}
envs = list_envs()
print({n: m for n, m in envs.items()
       if not n.startswith('_') and need <= set(m.get('packages', []))})
# 

r = build_env('proteomics_gpu', hydrate=True)
print(r['image'], r['spec_sha'], r['volumes'])
```

--------------------------------

### Launch Viewer with Static Output

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/skill-creator/SKILL.md

Use this option to generate a static HTML file instead of launching a server, suitable for environments without a display.

```bash
nohup python <skill-creator-path>/eval-viewer/generate_review.py \
  <workspace>/iteration-N \
  --skill-name "my-skill" \
  --benchmark <workspace>/iteration-N/benchmark.json \
  --static <output_path> \
  > /dev/null 2>&1 &
VIEWER_PID=$!
```

--------------------------------

### Launch Evaluation Viewer

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/skill-creator/SKILL.md

Launch the evaluation viewer to review skill outputs and benchmark data. Pass the previous workspace for comparison in later iterations.

```bash
nohup python <skill-creator-path>/eval-viewer/generate_review.py \
  <workspace>/iteration-N \
  --skill-name "my-skill" \
  --benchmark <workspace>/iteration-N/benchmark.json \
  > /dev/null 2>&1 &
VIEWER_PID=$!
```

--------------------------------

### Python Kernel Plugin Example

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/skill-creator/SKILL.md

Demonstrates a minimal kernel.py file with a pandas import at module scope and a function that defers a non-starter-set import to its body. Ensure imports not in the starter set are deferred to function bodies to avoid load errors.

```python
# kernel.py
import pandas as pd  # starter-set package — OK at module scope

def annotate_df(df: pd.DataFrame, gene_col: str = "gene") -> pd.DataFrame:
    """Attach HGNC symbols; see SKILL.md ## Workflow step 3."""
    import requests  # not in starter set — defer to function body
    ...
    return df
```

--------------------------------

### Install Packages in Environment

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/customize/SKILL.md

Install additional packages into an existing conda environment if needed.

```python
manage_packages(mode="install", environment="<slug>", packages=[...])
```

--------------------------------

### Social Media Attribution Example

Source: https://github.com/hughyau/academicforge/blob/site-first/ATTRIBUTIONS.md

An example of how to attribute the use of Academic Forge on social media platforms.

```plaintext
Writing my paper with @ClaudeAI and Academic Forge - amazing integration
of skills from @k-dense-ai and @orchestra-research! 🎓
```

--------------------------------

### Create and Configure Compute Environment

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/remote-compute-modal/SKILL.md

Creates a Modal compute environment with specified resources like image, environment, GPU, CPU, memory, and volumes. This is the first step before submitting a job.

```python
# repl tool — cell ① submits and RETURNS
c = host.compute.create('byoc:modal', provider_params={'modal': {
    'image':   r['image'],               # 'im-…' from build_env(); bare-name fallback exists but is brittle
    'env':     'proteomics_jax_gpu',     # bundled-env NAME → its shipped egress declaration applies
    'gpu':     'A100-40GB',
    'cpu':     8,
    'memory':  32768,                    # MiB — Modal's unit, not GiB
    'volumes': {'/cache': 'af2-params'}, # mount path → Modal Volume name
    'timeout': 3600,                     # container lifetime, seconds — optional; omitted → Settings default; ceiling: Modal's 24 h platform limit
}})
# → tier card → Sandbox.create() under the user's workspace → handle bound to sandbox_id

job = c.submit_job(
    intent='colabfold on target.fasta — 1× A100, ~40 min',
    command='colabfold_batch target.fasta out/ --num-recycle 3 --model-type alphafold2_ptm',
    inputs=[{'src': 'target.fasta', 'dst_filename': 'target.fasta'}],
    outputs=['out/ranked_0.pdb', {'glob': 'out/*.json', 'visibility': 'featured'},
             {'glob': '*.log', 'visibility': 'hidden'}],
    timeout_seconds=2700,                # job guard — optional; default = remaining container life − harvest margin
)
print('JOB_ID:', job.job_id)   # ← cell ends here
```

--------------------------------

### Academic Paper Attribution Example

Source: https://github.com/hughyau/academicforge/blob/site-first/ATTRIBUTIONS.md

An example of how to cite the Academic Forge in academic papers, mentioning its integration of skills from various sources.

```plaintext
We used the Academic Forge skill collection for Claude Code
(https://github.com/HughYau/academic-forge), which integrates
skills from k-dense-ai, orchestra-research, blader, and op7418.
```

--------------------------------

### Download LigandMPNN Model Parameters

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/compute-env-setup/references/envs_reference.md

Navigates to the LigandMPNN directory and executes a script to download the required model checkpoints.

```bash
cd /app/ligandmpnn && bash get_model_params.sh ./model_params
```

--------------------------------

### Initialize and Render Benchmark

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/skill-creator/eval-viewer/viewer.html

This is the main execution block that calls the initialization function and then the renderBenchmark function to display the results on the page.

```javascript
// ---- Start ---- init(); renderBenchmark();
```

--------------------------------

### Install Core DiffDock Python Packages

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/compute-env-setup/references/envs_reference.md

Installs essential Python packages for DiffDock, including RDKit, SciPy, and specific versions of e3nn and transformers.

```bash
pip install rdkit scipy networkx biopython<2 biopandas e3nn==0.5.1 spyrmsd pyyaml pandas prody fair-esm transformers<4.51 accelerate huggingface-hub gradio requests
```

--------------------------------

### Install PyTorch Geometric Dependencies for DiffDock

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/compute-env-setup/references/envs_reference.md

Installs PyTorch Geometric and its related libraries using specific find_links to ensure compatibility with the PyTorch version.

```bash
pip install torch_geometric
pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv --find-links https://data.pyg.org/whl/torch-2.4.0+cu124.html
```

--------------------------------

### Install X11 Render Libraries for RDKit

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/compute-env-setup/references/envs_reference.md

Installs necessary X11 libraries required by RDKit's Chem.Draw functionality, which is a dependency of pdbeccdutils used by openfold3.

```bash
apt-get install -y libxrender1 libxext6 libsm6
```

--------------------------------

### Package Skill for Distribution

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/skill-creator/SKILL.md

Package the skill into a .skill bundle for environments that do not use the `present_files` tool. This command should only be run if you have access to the `present_files` tool.

```bash
python -m scripts.package_skill <path/to/skill-folder>
```

--------------------------------

### Get Download URI Utility

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/skill-creator/eval-viewer/viewer.html

A utility function to get the download URI for a file. This is a placeholder and likely needs implementation based on how files are stored or generated.

```javascript
function getDownloadUri(file) { if (file.data_uri) return
```

--------------------------------

### Install Pip Packages for Proteomics

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/compute-env-setup/references/envs_reference.md

Installs a series of Python packages in a specific order to manage dependencies and resolve conflicts for proteomics tools like chai, boltz, cuequivariance, and openfold.

```bash
pip install chai_lab huggingface-hub biopython==1.84
pip install torch==2.7.1 --force-reinstall --extra-index-url https://download.pytorch.org/whl/cu128
pip install boltz cuequivariance-torch>=0.1.0 cuequivariance fair-esm deepchem[torch] transformers<5 pytorch-lightning prody>=2.5 biotite mdtraj
pip install cuequivariance-ops-torch-cu12
pip install ml-collections ihm modelcif
pip install protenix --no-deps
pip install openfold3==0.4.1 --no-deps
pip install rdkit pdbeccdutils kalign-python lmdb ijson func_timeout memory_profiler click wandb boto3 awscrt nvidia-cutlass
```

--------------------------------

### Create and Submit Compute Job

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/remote-compute-modal/SKILL.md

Create a compute instance using a previously built environment and submit a job to it. This involves specifying the image, environment name, GPU type, and volumes, then defining the command to execute.

```python
# repl tool 
import json
r = json.load(open('built_env.json'))
c = host.compute.create('byoc:modal', provider_params={'modal': {
    'image': r['image'],          # the im-… ref (bare name fallback brittle)
    'env': 'proteomics_gpu',      # bundled-env NAME 
    'gpu': 'A100',
    'volumes': r['volumes'],
}})
j = c.submit_job(
    intent='load esm2_t6 on GPU',
    command='python -c "from transformers import EsmModel; import torch; ' 
            'm = EsmModel.from_pretrained(\"facebook/esm2_t6_8M_UR50D\").cuda(); ' 
            'print(torch.cuda.get_device_name())"',
    timeout_seconds=300,
)
print('JOB_ID:', j.job_id)
```

--------------------------------

### Install JAX with CUDA 12 Support

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/compute-env-setup/references/envs_reference.md

Installs JAX with CUDA 12 support, ensuring compatibility with Hopper GPUs by upgrading jaxlib to a build with sm_90 PTX. This is crucial for avoiding CUDA_ERROR_INVALID_IMAGE errors.

```bash
pip install "jax[cuda12]>=0.4.26,<0.5" numpy<2 dm-haiku<0.0.13
```

--------------------------------

### Initialize Compute Connection

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/remote-compute-ssh/SKILL.md

Establish a connection to the compute provider. This is typically done once before submitting jobs.

```python
c = host.compute.create(provider)
```

--------------------------------

### List Connectors with Python SDK

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/customize/SKILL.md

Use `host.agents.list_connectors()` to retrieve information about available connectors. Without an argument, it lists all connectors. With a `connector_name`, it returns details for a specific connector, including its available tools.

```python
host.agents.list_connectors(connector_name=None)
# no arg → [{"name", "displayName", "source", "description",
#            "authState", "attachedAgents": [...]}, ...]
# with connector_name → single dict with an extra
#   "tools": [{"name", "description"}, ...] 
```

--------------------------------

### Install ESMFold2 Dependencies with uv

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/esmfold2/SKILL.md

Installs core dependencies including PyTorch, einops, rdkit, and ESMFold2 from GitHub. Ensure you have CUDA 12.x and Python 3.12. This command uses `uv` for efficient package management.

```bash
pip install --no-cache-dir uv
uv venv --python 3.12 /work/venv && source /work/venv/bin/activate
uv pip install \
  "torch>=2.5,<2.8" einops "biotite>=1.0" rdkit msgpack-numpy biopython \
  scikit-learn brotli attrs pandas cloudpathlib httpx tenacity zstd pydssp \
  pygtrie accelerate huggingface_hub safetensors "numpy<3" networkx \
  sentencepiece tokenizers regex packaging filelock pyyaml typing_extensions \
  "transformers @ git+https://github.com/Biohub/transformers.git@3a8956fb4d4ea16b0ec8e71deef2c2909b6a5cbf"
uv pip install --no-deps "esm @ git+https://github.com/Biohub/esm.git@f652b471"
# OPTIONAL — only affects ESMC attention; trunk speedup comes from set_kernel_backend("fused")
uv pip install ninja packaging wheel setuptools
MAX_JOBS=8 uv pip install --no-deps --no-build-isolation "flash-attn<3"
# Do NOT install transformer-engine — RuntimeError (not ImportError) on import
# slips ESMC's guard and kills ESMFold2Model import.
```

--------------------------------

### Load Skill Creator

Source: https://github.com/hughyau/academicforge/blob/site-first/skills/claude-science/customize/SKILL.md

Loads the skill creator tool. This is the first step before authoring any skill.

```python
skill({"skill": "skill-creator"})
```