AGENT SKILL

target-preparation

Validate a BindCraft target PDB and construct a target JSON with explicit chains, hotspots, binder length range, naming, and output-path checks.

License 许可
MIT
最近更新
2026-09-03
Stars
★ 351
Forks
27

01 前言 Front Matter

name
target-preparation
license
MIT
metadata
{ "disco-role": "operating" }
description
Validate a BindCraft target PDB and construct a target JSON with explicit chains, hotspots, binder length range, naming, and output-path checks.
disable-model-invocation
true

02 资源文件 3

03 SKILL.md 正文

Prepare a BindCraft target

Use this sub-skill before launching a design campaign. It turns a target PDB and design intent into a checked, portable target-settings JSON; it does not run AF2, ProteinMPNN, PyRosetta, or a design trajectory.

Route

  1. Keep the target PDB trimmed to the smallest biologically justified region. Preserve the target chains and standard amino-acid residues needed for the intended interface; do not assume that trimming is harmless for a target-dependent result.

  2. Copy the seven-key template and replace every path with a path valid on the machine that will launch BindCraft. Do not make the checked-in example's absolute paths a runtime dependency.

  3. Set chains to the target chain IDs. Use explicit chain-qualified hotspots for a multi-chain target when residue numbering could be ambiguous.

  4. Set target_hotspot_residues to null for AF2-selected site discovery, or to the documented residue/range syntax for a deliberate site. Validate every selected chain and hotspot against the PDB when the PDB is available.

  5. Run the bundled read-only validator:

    python skills/disco/bindcraft/sub-skills/target-preparation/scripts/validate_target.py \
      --target-json ./settings_target/my_target.json \
      --pdb ./inputs/my_target.pdb
    

    Omitting --pdb still validates JSON and reports declared paths, but cannot prove that chains or residue numbers exist. The validator never creates, rewrites, downloads, or imports from the BindCraft checkout.

  6. Hand the validated JSON to the design route. For GPU/model, algorithm, filter, and launch decisions, see design-pipeline. For post-run artifacts and metrics, see results-analysis.

Schema and syntax details are in data-formats; input and filesystem failure recovery is in troubleshooting.

Safety boundary

The JSON/PDB checks are CPU-safe inspection only. BindCraft's actual AF2/MPNN binder-design loop requires the prepared CUDA environment, model weights, and other documented prerequisites; a successful target validation is not evidence that a design can run or that a target will yield binders. PyRosetta licensing, AF2-weight availability, GPU memory, and target-dependent behavior remain separate launch gates.

04 同仓库更多 Skills

VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-triton-opt

Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-solve

END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-skill

PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-pattern-fusion

Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-orchestrate

Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-feedback

Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).

★ 351