Date: 2026-03-20 Decision: Use the proven 7-phase research-first methodology from Blender, ERPNext, and Draw.io packages. Rationale: This approach produces higher-quality, deterministic skills because research precedes creation. Proven across 4 prior packages with 150+ skills total.
Date: 2026-03-20 Decision: Evaluate existing QGIS MCP servers before deciding build vs. configure. Rationale: The QGIS ecosystem may have existing MCP integrations. Evaluate before committing to a custom build. Decision will be revisited in Phase B0.5.
Date: 2026-03-20 Decision: Target ~19 skills across 5 categories (core: 3, syntax: 4, impl: 8, errors: 2, agents: 2). Rationale: Based on initial QGIS technology landscape analysis. Covers the primary GIS workflows: vector/raster analysis, map production, data integration, and spatial data management. Will be refined in Phase B3 after deep research.
Date: 2026-03-20 Decision: Focus on the Claude ecosystem. No explicit agentskills.io open standard compliance. Rationale: Claude-specific features (allowed-tools, plugin marketplace, progressive disclosure) can be used freely without cross-platform constraints. Same decision as Draw.io package (D-017).
Date: 2026-03-20
Decision: Do NOT build a custom QGIS MCP server. Reference existing servers in skills instead.
Rationale: The ecosystem already has strong options: jjsantos01/qgis_mcp (855★, 15 tools, Beta, listed in official QGIS Plugin Repository) and nkarasiak/qgis-mcp (51 tools, actively developed, supports QGIS 3.28-4.x). Our skill package is complementary (knowledge vs runtime). For v2.0, consider contributing PRs to existing servers rather than building from scratch.
Date: 2026-03-20
Decision: Target QGIS 3.44 LTR (current stable) as primary version. Include QGIS 4.0 (Qt6) compatibility notes where relevant.
Rationale: QGIS 4.0 released 2026-03-06 but is brand new. The LTR (3.44) is what production users run. PyQGIS API is largely preserved in 4.0 but Qt6 changes affect some patterns (e.g., QMetaType.Type instead of QVariant.Type).
Date: 2026-03-20 Decision: Proceed with full 19-skill package. No existing QGIS/PyQGIS skills found in the Claude ecosystem. Rationale: Searched Claude Skills Hub, anthropics/skills repo, GitHub, OpenAEC Foundation, and community repos. Zero QGIS skills found. Closest related: K-Dense scientific skills (generic geospatial, no PyQGIS), gis-claude-skills (Leaflet only). We have a completely clear field.