feat: improve skill scores for 5 skills + add Tessl review workflow

.github/workflows/skill-review.yml

@@ -0,0 +1,22 @@
+# Tessl Skill Review — runs on PRs that change any SKILL.md; posts scores as one PR comment.
+# Docs: https://github.com/tesslio/skill-review
+name: Tessl Skill Review
+
+on:
+  pull_request:
+    branches: [main]
+    paths:
+      - "**/SKILL.md"
+
+jobs:
+  review:
+    runs-on: ubuntu-latest
+    permissions:
+      pull-requests: write
+      contents: read
+    steps:
+      - uses: actions/checkout@v4
+      - uses: tesslio/skill-review@main
+      # Optional quality gate (off by default — do not enable unless user asked):
+      # with:
+      #   fail-threshold: 70

skills/alphafold/SKILL.md

@@ -1,6 +1,12 @@
+---
+name: alphafold
+description: "Use when running AlphaFold2 predictions on custom protein sequences, validating designed sequences via self-consistency, predicting binder-target complexes, or interpreting AF2 confidence metrics (pLDDT, pTM, ipTM)."
+metadata:
+---
+
 # AlphaFold2 Structure Prediction
 
-Run AlphaFold2 for protein structure prediction and complex modeling. Use for validating designed sequences, predicting binder-target complexes, and calculating confidence metrics (pLDDT, pTM, ipTM).
+Use when the agent needs to run AlphaFold2 for protein structure prediction and complex modeling. Covers validating designed sequences, predicting binder-target complexes, and calculating confidence metrics (pLDDT, pTM, ipTM).
 
 Distinct from `alphafold-database` (which retrieves pre-computed structures) — this skill covers running AF2 predictions on custom sequences.
 

skills/boltz/SKILL.md

@@ -1,6 +1,12 @@
+---
+name: boltz
+description: "Use when predicting biomolecular structures (proteins, RNA, DNA, ligands) with the open-source Boltz diffusion model as an alternative to AlphaFold3."
+metadata:
+---
+
 # Boltz Structure Prediction
 
-Open-source biomolecular structure prediction using diffusion models. MSA-optional, handles proteins, RNA, DNA, small molecules, ions, and covalent modifications in a single model. Strong alternative to AlphaFold3.
+Predict biomolecular structures using Boltz, an open-source diffusion model. Boltz handles proteins, RNA, DNA, small molecules, ions, and covalent modifications in a single model without requiring multiple sequence alignments (MSA-optional). It serves as a strong open-source alternative to AlphaFold3.
 
 ## Requirements
 

skills/chai/SKILL.md

@@ -1,6 +1,12 @@
+---
+name: chai
+description: "Use when predicting molecular structures (proteins, nucleic acids, small molecules, and complexes) with the Chai-1 foundation model via local inference or the Chai Discovery API."
+metadata:
+---
+
 # Chai-1 Structure Prediction
 
-Foundation model for molecular structure prediction. Handles proteins, nucleic acids, small molecules, and complexes. Fast inference, strong multimer performance, available via API and local install.
+Use when the user needs to predict molecular structures — proteins, nucleic acids, small molecules, or multi-chain complexes — using the Chai-1 foundation model. Supports both local GPU inference and the Chai Discovery API for remote execution.
 
 ## Requirements
 
@@ -143,7 +149,7 @@ print(f"Interface PAE: {interface_pae:.2f} Å (< 10 = good)")
 
 ## Use Cases
 
-- **Fast validation**: Quickly predict binder-target complexes before expensive MD
-- **Ligand complexes**: Predict protein-small molecule binding poses
-- **Ensemble scoring**: Generate 5 models, rank by ipTM for design selection
-- **Nucleic acid interactions**: Protein-DNA/RNA complex prediction
+- **Fast validation**: Predicts binder-target complexes quickly before committing to expensive MD simulations.
+- **Ligand complexes**: Predicts protein-small molecule binding poses from SMILES input.
+- **Ensemble scoring**: Generates multiple models and ranks them by ipTM for design selection.
+- **Nucleic acid interactions**: Predicts protein-DNA/RNA complex structures.