Skills and connector packaging
Connector boundary
Section titled “Connector boundary”Bridge Town’s remote Claude and OpenAI directory profiles are static product catalogs. They contain tools, schemas, annotations, and safe product-reference metadata. They do not deliver behavioral prompt templates, user-authored instructions, or a runtime skill lookup surface.
The profile contracts make this explicit:
- Claude profiles have empty
resourcesandpromptsarrays. - OpenAI profiles have empty
resources,prompts, andskillsarrays until the future static importer is implemented. search_toolsreturns only ranked product tools. It never returns a skill match or a follow-up action that loads external behavior.
This boundary keeps a connector description or tool result from asking its host to fetch, render, honor, or follow dynamic instructions.
Separate packages
Section titled “Separate packages”The 19 Markdown files under the repository’s skills/ directory are source
material for separately distributed Claude skills/plugins. They are not included
as resources in the remote connector. A Claude Code or Claude Desktop package
may ship the files through its normal package mechanism, subject to its own
review and versioning.
The OpenAI/ChatGPT profile does not import these files. A later OpenAI-specific implementation may publish no more than five deliberately selected, immutable skills through OpenAI’s static importer snapshot. That importer owns its manifests, resource digests, limits, and rescan process; it is not a server-side alias for the repository’s Markdown files.
The public Skills REST directory is also separate. Its user-authored records must not be copied into a connector profile or treated as host instructions.
Static reference resources
Section titled “Static reference resources”Resources remain appropriate when they are immutable product data that cannot change host behavior. Examples include the native-sheet JSON schema and worked schema examples, dashboard chart-type schemas, and enum/reference tables. These resources describe data contracts; they do not contain agent workflow instructions.
Package catalog
Section titled “Package catalog”Every package has a valid frontmatter name/category/context declaration and its
related_tools names resolve against the canonical MCP tool manifest.
| Package file | Skill name | Category | Status |
|---|---|---|---|
bridge_town_gsheet_formatting.md | bridge-town-gsheet-formatting | Google Sheets | Model-output publishing is distinguished from dedicated value-writing and formatting surfaces. |
bridge_town_native_sheets.md | bridge-town-native-sheets | Native sheets | Current; references schema/reference resources and typed sheet tools. |
bridge_town_scenario_sheets.md | bridge-town-scenario-sheets | Native sheets | Current; branch/edit/run/compare workflow is complete. |
collaborate_on_project.md | collaborate_on_project | Collaboration | Current; workspace model templates are distinguished from skill packages. |
commit_files.md | commit_files | File management | Current; atomic file-commit guidance is complete. |
connect_google_sheets.md | connect_google_sheets | Data integration | Current; OAuth prerequisite and refresh path are explicit. |
convert_excel_workbook.md | convert_excel_workbook | Data integration | Uses the current workbook conversion tools and package metadata. |
create_cash_flow_model.md | create_cash_flow_model | Model authoring | Current; runnable model and entrypoint guidance is complete. |
create_dashboard.md | create_dashboard | Dashboards | Current; chart schemas and create/share flow are explicit. |
create_expense_model.md | create_expense_model | Model authoring | Current; required files and run path are complete. |
create_headcount_model.md | create_headcount_model | Model authoring | Current; required assumptions and run path are complete. |
create_marketplace_gmv_model.md | create_marketplace_gmv_model | Model authoring | Current; supply, demand, and take-rate assumptions are complete. |
create_revenue_model.md | create_revenue_model | Model authoring | Current; drivers, outputs, and run path are complete. |
create_unit_economics_model.md | create_unit_economics_model | Model authoring | Current; CAC/LTV assumptions and outputs are complete. |
create_usage_based_pricing_model.md | create_usage_based_pricing_model | Model authoring | Current; usage and pricing assumptions are complete. |
discover_and_query_data.md | discover_and_query_data | Data integration | Current; source discovery, freshness, and query guidance are complete. |
gsheet_finance_styles.md | gsheet-finance-styles | Google Sheets | Corrected: points to the canonical formatting package name and preserves the write/format split. |
run_scenario_analysis.md | run_scenario_analysis | Scenario analysis | Current; branch-based what-if workflow is complete. |
scenario_analysis.md | scenario_analysis | Model authoring | Current; new-model scenario authoring is distinct from branch analysis. |
This catalog does not make these files remotely discoverable. Packaging changes must update the relevant Claude plugin or future OpenAI importer snapshot and its digest evidence separately.