SandBase Skills v0.3.5 — verifiable source identity across Agent Skills hosts
This GitHub source release strengthens the portable multi-source-search workflow and makes its runtime requirements explicit.
What changed
- Canonical source URL identity: the offline evidence-ledger validator now rejects cosmetic duplicates caused by host/scheme case, default ports, fragments, and common tracking parameters. Meaningful query parameters remain distinct.
- Standard compatibility metadata:
multi-source-searchdeclares the host capabilities it needs through the Agent Skillscompatibilityfield. - Portable Agent Plugin: the canonical research Skill, references, and offline validator can be installed by Copilot CLI and other Agent Plugins 1.0 clients.
- Broader native installation: documented GitHub CLI, DeepSeek Harness, Codex, Claude Code, Cursor, and Gemini CLI paths.
- Discovery improvements: security-scanned directory links, install-count badge, localized calls to action, and a public Skill request form.
Try the flagship Skill
npx skills add sandbaseai/sandbase-skills@multi-source-searchOr preview it with GitHub CLI 2.90+:
gh skill preview sandbaseai/sandbase-skills research/multi-source-searchValidate a committed evidence ledger without an account or network request:
python3 research/multi-source-search/scripts/validate_report.py \
examples/verifiable-research-report.jsonVerification
- 29 automated tests
- 25 public SandBase registry Skills validated
- 88 Claude Code marketplace Skills validated
- Agent Plugin canonical-copy check
- 200-file package dry run
This is a GitHub source release. No npm package is published by this release.
Full comparison: v0.3.4...v0.3.5