mapbox__mapbox-agent-skills

skills/mapbox__mapbox-agent-skills

Fork 0

mirror of https://github.com/mapbox/mapbox-agent-skills.git synced 2026-09-14 20:16:39 +08:00

Commit Graph

Author	SHA1	Message	Date
Matthew Podwysocki	c482cc8c34	Fix eval.js default model: claude-sonnet-4-20250514 has been retired (#74 ) npm run eval was failing with a 404 not_found_error for anyone not overriding EVAL_MODEL/EVAL_JUDGE_MODEL, since the hardcoded default model ID is no longer served by the API. Default to claude-sonnet-5 instead.	2026-08-05 09:27:26 -04:00
Mofei Zhu	65bab68c18	Improve eval scores: fix skill content for 5 skills to reach 100% - mapbox-mcp-runtime-patterns: Use MCPServerHTTP for pydantic-ai, proper MCP client/transport for LangChain, add directions_tool parameter examples - mapbox-store-locator-patterns: Add marker strategy table with WebGL/GPU explanation, clarify GeolocateControl vs getCurrentPosition separation - mapbox-search-integration: Add session token guidance for direct API usage - mapbox-maplibre-migration: Emphasize all MapLibre plugins need Mapbox equivalents - mapbox-geospatial-operations: Strengthen optimization_tool vs directions/matrix distinction, add search_and_geocode_tool as pre-step - mapbox-web-performance-patterns: Fix clustering threshold to 10,000+ - scripts/eval.js: Auto-load .env file	2026-03-30 16:41:50 +03:00
Mofei Zhu	7cac917a48	Split 16 skills into slim SKILL.md + references/ for progressive disclosure Per Agent Skills Specification, move detailed content to references/ files that agents load on demand. Total SKILL.md lines: 15,554 → 5,088 (-67%). Skills split: search-integration (9 refs), mcp-runtime-patterns (8), web-integration-patterns (8), style-patterns (7), data-visualization (6), store-locator (6), google-maps-migration (5), android-patterns (5), web-performance (5), ios-patterns (5), mcp-devkit-patterns (4), search-patterns (3), maplibre-migration (3), style-quality (3), token-security (2), cartography (2). Total: 81 reference files. Only geospatial-operations not split (decision framework too interconnected). Also adds: - eval.js script and baseline eval results - Updated .gitignore for eval workspaces - package.json version bump and eval scripts	2026-03-30 15:29:32 +03:00

Author

SHA1

Message

Date

Matthew Podwysocki

c482cc8c34

Fix eval.js default model: claude-sonnet-4-20250514 has been retired (#74 )

npm run eval was failing with a 404 not_found_error for anyone not
overriding EVAL_MODEL/EVAL_JUDGE_MODEL, since the hardcoded default model
ID is no longer served by the API. Default to claude-sonnet-5 instead.

2026-08-05 09:27:26 -04:00

Mofei Zhu

65bab68c18

Improve eval scores: fix skill content for 5 skills to reach 100%

- mapbox-mcp-runtime-patterns: Use MCPServerHTTP for pydantic-ai, proper MCP
  client/transport for LangChain, add directions_tool parameter examples
- mapbox-store-locator-patterns: Add marker strategy table with WebGL/GPU
  explanation, clarify GeolocateControl vs getCurrentPosition separation
- mapbox-search-integration: Add session token guidance for direct API usage
- mapbox-maplibre-migration: Emphasize all MapLibre plugins need Mapbox equivalents
- mapbox-geospatial-operations: Strengthen optimization_tool vs directions/matrix
  distinction, add search_and_geocode_tool as pre-step
- mapbox-web-performance-patterns: Fix clustering threshold to 10,000+
- scripts/eval.js: Auto-load .env file

2026-03-30 16:41:50 +03:00

Mofei Zhu

7cac917a48

Split 16 skills into slim SKILL.md + references/ for progressive disclosure

Per Agent Skills Specification, move detailed content to references/ files
that agents load on demand. Total SKILL.md lines: 15,554 → 5,088 (-67%).

Skills split: search-integration (9 refs), mcp-runtime-patterns (8),
web-integration-patterns (8), style-patterns (7), data-visualization (6),
store-locator (6), google-maps-migration (5), android-patterns (5),
web-performance (5), ios-patterns (5), mcp-devkit-patterns (4),
search-patterns (3), maplibre-migration (3), style-quality (3),
token-security (2), cartography (2). Total: 81 reference files.

Only geospatial-operations not split (decision framework too interconnected).

Also adds:
- eval.js script and baseline eval results
- Updated .gitignore for eval workspaces
- package.json version bump and eval scripts

2026-03-30 15:29:32 +03:00

3 Commits