marshalleriksen-neura/deeting/provider-doc-ingestion

Provider Doc Ingestion

Overview

Official skill for collecting provider documentation pages, extracting source-linked protocol evidence, and drafting verify-ready provider preset candidates.

• Runtime: local
• Access: all
• Permissions: network.outbound

This skill is the ingestion lane. It does not publish provider presets by itself.

Available Tools

• collect_provider_doc_evidence — Fetch one or more provider documentation pages and return raw evidence items plus any fetch failures.
• draft_provider_candidate — Convert a structured extraction report into a provider preset candidate with verification gaps preserved and a ready-to-use provider_registry_handoff payload.
• score_provider_candidate_readiness — Report whether the drafted candidate is ready for schema alignment and dry-run verification.
• build_provider_registry_handoff — Rebuild the explicit downstream tool payloads from a candidate if AI or UI needs a clean final handoff object.

Usage Notes

• Use this skill before provider_registry when the source of truth is a provider's public documentation.
• Keep explicit and inferred evidence separate.
• Do not publish directly from this package.
• collect_provider_doc_evidence also returns:
• protocol_profile_template
• provider_registry_handoff_template
• draft_provider_candidate returns:
• protocol_profiles
• provider_registry_handoff
• build_provider_registry_handoff returns:
• handoff.get_unified_schema
• handoff.verify_provider_template
• handoff.save_local_provider_preset
• The intended downstream flow is:

1. collect_provider_doc_evidence
2. external or model-assisted extraction into ProviderExtractionReport
3. draft_provider_candidate
4. score_provider_candidate_readiness
5. handoff to provider_registry.get_unified_schema
6. handoff to provider_registry.verify_provider_template
7. write to the local provider market file only after verification succeeds