AI Lab for Cities at Harvard
Official@AI-Lab-for-Cities-at-Harvard
Harvard AI Lab for Cities publishes MIT/CC0 civic skills for government performance management, policy analysis, plain-language notices, and portable skill localization.
Agent Skills by AI Lab for Cities at Harvard
Showing 4 vetted skills indexed across 1 GitHub repositories.
jurisdiction-priority-scan
Scans open data, media, and testimony to rank agency priority problems.
localize-skill
Fills a generalized skill's context slots with one organization's confirmed local values.
generalize-skill
Extracts organization-specific values from a skill into a portable generalized skill and context file.
plain-language-notice-rewriter
Rewrites government notice templates into plain language at a target reading level.
Frequently Asked Questions About AI Lab for Cities at Harvard
FAQPage SchemaWhat tasks can I accomplish with AI Lab for Cities at Harvard skills?▼
You can scan a jurisdiction's performance data, media, and testimony to rank agency problems; orchestrate full civic policy analyses with peer benchmarking; rewrite government notice templates to plain-language reading levels; and generalize or localize skills so they transfer between organizations.
Who are these civic skills designed for?▼
They target US state and city performance officers, agency leaders, municipal analysts, and civic innovation teams working on 311, health and human services, labor, emergency dispatch, housing, safety, or transportation. Skill authors sharing work across jurisdictions also use the generalize/localize pair.
What are the installation and runtime requirements?▼
jurisdiction-priority-scan requires an OpenContext MCP connector to the jurisdiction's open data portal plus web search and fetch access; social media access is optional. The generalize, localize, and plain-language skills run offline, reading and writing local files only, with no credentials or external services.
Are these skills open source and free to use?▼
Yes. jurisdiction-priority-scan and civic-analytics-workflow are MIT licensed, while localize-skill, generalize-skill, and plain-language-notice-rewriter are CC0-1.0 public domain. All are free, with no deployment requirements and no data sensitivity concerns noted in their manifests.
What limitations or prerequisites should I know before using them?▼
The priority scan assumes US federal programs and requires an open data connector, degrading without one. The notice rewriter handles templates only, never filled notices with personal data, and preserves legal elements you specify. All outputs are drafts requiring human review and approval.