AI Engineering

ProteinMPNN designs amino acid sequences for a given protein backbone using inverse folding. Use it for redesigning structures, fixing selected residues, or generating variants for expression, stability, and multi-state design.

RFDiffusion generates protein backbones for de novo design, binder scaffolds, motif scaffolding, and symmetric oligomers. Use it when you need a backbone to pass to proteinmpnn, alphafold, chai, or protein-qc.

SolubleMPNN designs protein sequences with a soluble model variant for E. coli expression and higher solubility. Use it to reduce aggregation or inclusion bodies when standard ProteinMPNN is too hydrophobic.

Tool Schema Design helps define model-facing tools with clear names, bounded JSON Schema parameters, explicit side effects, and safe defaults. Use it when building function-calling or MCP tools and when models choose the wrong tool or invent arguments.

UniProt retrieves protein sequences, annotations, domain boundaries, variants, and cross-references from the UniProt REST API. Use it to look up proteins by accession or gene name and to find related PDB structures.

Audit Plan evaluates an incoming project plan against bigpowers principles and conventions, then records gaps and a READY or NOT READY verdict before build work starts. Use it when a new project arrives or before seeding conventions in an unfamiliar codebase.

Context7 MCP fetches current library documentation through Context7 instead of relying on training data. Use it for framework, API, setup, and example questions about libraries like React, Next.js, or Prisma.

Craft Skill helps create new Bigpowers skills with the required structure, triggers, and bundled resources. Use when you need to write, build, or package a new skill for the Bigpowers lifecycle.

Craft Skill helps create new bigpowers skills with proper structure, progressive disclosure, and bundled resources. Use when you need to write or build a new skill for the bigpowers lifecycle.

Design Interface generates multiple radically different module interfaces in parallel and compares their trade-offs before choosing one. Use it when designing an API, exploring module shapes, or following a “design it twice” approach.

Dispatch Agents runs multiple subagents in parallel on independent tasks. Use it when work can be split into decoupled pieces and speed matters, with explicit synchronization points and no inter-task review gate.

LLM Council wraps pr-council-review with external CLI coding agents for an independent second opinion before accepting a GREEN verdict. Use it for high-stakes, security-sensitive, or architecturally significant reviews.