in-silico

Operates the EBFC Gen 2.0 DFT and MD computational pipeline with cache-discipline rules.

23|1|Updated Feb 26, 2026
One-click install
npx skills add https://github.com/Alexey-Lukin/silken_net --skill in-silico-alexey-lukin
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: in-silico
Source: https://github.com/Alexey-Lukin/silken_net/tree/main/.claude/skills/in-silico
Command: npx skills add https://github.com/Alexey-Lukin/silken_net --skill in-silico-alexey-lukin

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Running and modifying a large DFT/MD computational chemistry pipeline (AlphaFold-3, OpenMM, PySCF) without breaking downstream caches, verdicts, or documentation is error-prone; this Skill encodes the dependency graph, hard-won DFT/MD gotchas, and modification discipline so changes propagate correctly. ## Core Features & Use Cases - Script Dependency Graph: Maps which L1-L4 and anchor-mechanics scripts read each other's caches, so a change re-runs exactly the downstream set. - Critical Rules & Gotchas: Documents PySCF heavy-metal pitfalls (no density_fit for Os/Ce, level_shift for oscillating SCF), MD NaN prevention (10K pre-relax, 50K ramp), and PCET solvation traps. - Cache-as-SSOT Discipline: Enforces one-cache-per-model ownership, doc-vs-cache verification, and field-by-field diffing when regenerating results. - Use Case: When the Os(III) SCF oscillates forever or the FAD-to-Os redox cascade computes uphill, consult the gotchas and cascade verdict routing instead of re-deriving known method limits. ## Quick Start Ask the assistant to run or modify an in-silico DFT or MD script, for example to add a ligand to the pipeline or diagnose why a PySCF calculation fails to converge.

Frequently Asked Questions about in-silico

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I add a new ligand to the MD pipeline?

Adding a ligand is a five-step chain: parameterize it in a 0N script, add it to the GAFF cache, place the SDF in ligands/, add a test, then rerun all downstream MD that uses it. Skipping the last step leaves downstream MD stale.

Why does the Os(III) SCF calculation oscillate forever in PySCF?

Open-shell transition metals can oscillate indefinitely; setting level_shift=0.3 stops UKS oscillation on Os(III), Co-Ce, and Ce. It biases the reported LUMO, so enable it only where the SCF oscillates and read E_total instead of LUMO.

Why does OpenMM MD go NaN during the NVT temperature ramp?

MD goes NaN at the NVT ramp unless the system pre-relaxes at 10K and ramps from 50K in 10K steps, with maxIterations=10000 for minimization. In CI smoke tests the cause was box size, not RNG, so shrink the box rather than re-running.

Should I use density fitting for heavy metals in PySCF?

No. For Os and Ce the auto-generated auxiliary basis makes density_fit about three times slower than standard integrals. Use lanl2dz for Cu/Co and stuttgart_rsc for Ce, since Ce is not covered by lanl2dz.

How do I verify a documentation number against computed results?

The cache JSON is ground truth; any computed number in a doc must match it. The pinned checks live in tools/in_silico/tests/test_doc_cache_sync.py as context-anchored regexes, and after a recompute you grep both the new and old values across all docs.

Why does the FAD-to-Os redox cascade compute uphill in DFT?

The cascade is verified downhill from measured redox potentials, while raw DFT comes out uphill in every method; this is a method limit, not physics. The gap decomposes into Os mediator speciation, differential solvation, and the 4,4'-dimethyl substituent.