| name | fold-cartographer |
| description | Use when interpreting auto-AlphaFold3 scorer diagnostics into exactly one Fold Cartographer failure target, preserving best_val_calpha_lddt as the primary objective and mapping local geometry, long-range topology, distogram-vs-3D, or stability/compute evidence to safe move-family suggestions. |
Fold Cartographer
Use this skill to convert scorer diagnostics into one routing target for the
next AlphaFold3-lite hypothesis. The output guides research; it does not change
the objective.
Workflow
- Read
references/routing.md.
- Inspect primary score and diagnostics.
- Choose exactly one diagnostic target.
- Name two or three compatible move families.
- Explain the evidence briefly.
Hard Rules
- The only primary objective is
best_val_calpha_lddt.
- Do not optimize diagnostics directly.
- Do not use hidden validation for search routing.
- Do not ask to inspect labels, scorer code, validation splits, or locked data.
- Do not call
modal run, Modal APIs, or submit trials.
Output Contract
TARGET: <one target>
EVIDENCE: <metrics that justify target>
MOVE_FAMILY_CANDIDATES: <comma-separated families>
AVOID: <moves that would violate the contract or chase diagnostics directly>
References
- Read
references/routing.md for target definitions and tie-break behavior.