jinzhezenggroup/computational-chemistry-agent-skills/dft-siesta

SIESTA Task Router

Use this skill as the top-level SIESTA orchestration layer.

Purpose

This skill routes requests to one task-specific SIESTA subskill path:

• dft-siesta/static
• dft-siesta/relax
• dft-siesta/md
• dft-siesta/electronic

Scope

This router skill should:

• require user-provided structure or prerequisite context
• classify request intent into one SIESTA task type
• collect minimal shared context before dispatch
• delegate detailed parameter handling to selected subskill
• enforce consistent output/handoff policy across subskills

This router skill should not:

• own detailed templates for all SIESTA tasks
• execute or submit calculations
• bypass subskill-specific guardrails

Hard requirement

The user must provide enough starting context:

• structure input for static / relax / md
• prerequisite converged context for electronic

If prerequisites are missing, stop and ask for them.

Routing rules

1. If user requests single-point SCF/energy: route to dft-siesta/static.
2. If user requests geometry optimization: route to dft-siesta/relax.
3. If user requests molecular dynamics: route to dft-siesta/md.
4. If user requests electronic analysis workflow: route to dft-siesta/electronic.
5. If intent is ambiguous, ask one focused clarification question before routing.

Shared policy for all subskills

• do not invent missing pseudopotential files
• expose assumptions and unresolved scientific choices explicitly
• return handoff-ready task layout
• if execution is requested, hand off to dpdisp-submit

Output from router

Provide:

1. selected subskill path
2. why it was selected
3. minimal missing inputs (if any)
4. explicit next step (invoke selected subskill)