Update to Forward Deployed

Changed files

Choose a file to see its text changes. Binary or large files show metadata only.

PRIVACY.md →

README.md →

docs/installation.md →

mcp.json →

plugin.json →

skills/adoption-and-scale/SKILL.md →

skills/capability-exploration/SKILL.md →

skills/context-design/SKILL.md →

skills/enterprise-integration/SKILL.md →

skills/evaluate-and-improve/SKILL.md →

skills/get-started/SKILL.md →

skills/opportunity-design/SKILL.md →

skills/use-case-portfolio/SKILL.md →

skills/value-case/SKILL.md →

PRIVACY.md

− removed · + added

--- before
+++ after
@@ -1,24 +1,24 @@
-# Privacy
-
-The Forward Deployed skills are local instructions. Using a skill does not by itself send data to the publisher. Your AI host and any connected services handle the work under their own permissions and settings.
-
-The optional workspace at `forward-deployed.unsol.dev` processes the information you send to its tools. Stateless mode is the default: the service receives a workspace snapshot, processes the request and returns it without saving the workspace in its database. Your host retains the returned snapshot according to its own settings. The website keeps its active snapshot in browser tab session storage, including through a sign-in redirect. Export work you want to keep.
-
-If you choose to save online, the service stores your workspace in a database managed by ChatGPT Sites, scoped to your signed-in account. This includes the source excerpts, opportunities, requirements, delivery work, scenarios and runs you choose to save. Sites manages ChatGPT sign-in. Forward Deployed uses the Site-scoped user ID to separate saved work and may display your profile name; it does not store your email or ChatGPT authentication tokens. Its public connector stores OAuth client registrations, consent transactions and access grants in the same managed database to authenticate connected hosts. Temporary authorization cookies protect that connection flow. Connecting the plugin does not itself save a workspace. The service does not automatically import meetings or write to a Git repository.
-
-Daily aggregate usage counters record fixed tool names, storage mode, result category, call count and total elapsed time. They exclude account IDs, IP addresses, project names, source content, URLs and request payloads. There is no browser analytics script or tracking cookie. You can turn usage reporting off in the workspace, pass `telemetry: false` to a tool, or send `DNT: 1`, `Sec-GPC: 1` or `x-fd-telemetry: off`. The application does not log request bodies. ChatGPT Sites and its infrastructure providers process operational data under their own policies to host and secure the service.
-
-Exports go to your chosen destination. A Git handoff prepares files for your host's Git tools; it does not grant repository access or push changes. Review an export before sharing it, especially if it includes meeting excerpts or client information.
-
-Information sent to the workspace is processed by Forward Deployed on ChatGPT Sites and its infrastructure, including Cloudflare. Your connected AI host receives tool results, and exports go to the destination you select. The maintainer may access stored records to operate the service or handle a support or removal request.
-
-## Retention and controls
-
-- **Stateless work:** the application processes each request without saving its workspace in the database. Browser copies remain in tab session storage; browser session restoration may preserve them. Clear the site's browser storage to remove local copies. Your AI host and export destinations have their own retention settings.
-- **Saved work:** workspaces and records used to make retries safe have no automatic expiry. Retry records can include earlier workspace content. They remain until the maintainer removes them, including in response to a verified removal request. There is currently no self-service delete control or fixed removal-response deadline. Disconnecting the plugin does not delete saved work.
-- **Connector authorization:** consent transactions and their temporary browser cookies expire after 10 minutes. Access tokens are valid for up to one hour; refresh grants expire 30 days after issuance. Expired database records cannot be used and are removed on a subsequent OAuth storage write, so expiry is not a guaranteed physical deletion time. Client registrations have no automatic expiry and remain until removed by the maintainer.
-- **Usage counts:** daily aggregate counters are retained indefinitely; no automatic deletion schedule is configured. Turning reporting off stops future opted-out measurements and does not remove earlier aggregate counts. The counters contain no account identifier with which to locate an individual's contribution.
-
-ChatGPT Sites manages its own sign-in, native connection records and operational data. Its providers' retention and backup timelines are governed by their policies; Forward Deployed does not set or promise a deletion period for those copies.
-
-You can keep work stateless, export it, clear local browser storage, turn usage reporting off, or disconnect the connector in your AI host. For privacy questions or removal requests, use the [repository](https://github.com/anotb/forward-deployed) to ask the maintainer for a private contact channel. Keep that initial request nonsensitive: do not include account identifiers, client material, credentials or other personal information in a public issue. Account verification and details of the records to remove should be shared only through the agreed private channel.
+# Privacy
+
+The Forward Deployed skills are local instructions. Using a skill does not by itself send data to the publisher. The Claude plugin loads only these skills and does not connect to the workspace described below. Your AI host and any connected services handle the work under their own permissions and settings.
+
+The optional workspace at `forward-deployed.unsol.dev` processes the information you send to its tools. Stateless mode is the default: the service receives a workspace snapshot, processes the request and returns it without saving the workspace in its database. Your host retains the returned snapshot according to its own settings. The website keeps its active snapshot in browser tab session storage, including through a sign-in redirect. Export work you want to keep.
+
+If you choose to save online, the service stores your workspace in a database managed by ChatGPT Sites, scoped to your signed-in account. This includes the source excerpts, opportunities, requirements, delivery work, scenarios and runs you choose to save. Sites manages ChatGPT sign-in. Forward Deployed uses the Site-scoped user ID to separate saved work and may display your profile name; it does not store your email or ChatGPT authentication tokens. Its public connector stores OAuth client registrations, consent transactions and access grants in the same managed database to authenticate connected hosts. Temporary authorization cookies protect that connection flow. Connecting the plugin does not itself save a workspace. The service does not automatically import meetings or write to a Git repository.
+
+Daily aggregate usage counters record fixed tool names, storage mode, result category, call count and total elapsed time. They exclude account IDs, IP addresses, project names, source content, URLs and request payloads. There is no browser analytics script or tracking cookie. You can turn usage reporting off in the workspace, pass `telemetry: false` to a tool, or send `DNT: 1`, `Sec-GPC: 1` or `x-fd-telemetry: off`. The application does not log request bodies. ChatGPT Sites and its infrastructure providers process operational data under their own policies to host and secure the service.
+
+Exports go to your chosen destination. A Git handoff prepares files for your host's Git tools; it does not grant repository access or push changes. Review an export before sharing it, especially if it includes meeting excerpts or client information.
+
+Information sent to the workspace is processed by Forward Deployed on ChatGPT Sites and its infrastructure, including Cloudflare. Your connected AI host receives tool results, and exports go to the destination you select. The maintainer may access stored records to operate the service or handle a support or removal request.
+
+## Retention and controls
+
+- **Stateless work:** the application processes each request without saving its workspace in the database. Browser copies remain in tab session storage; browser session restoration may preserve them. Clear the site's browser storage to remove local copies. Your AI host and export destinations have their own retention settings.
+- **Saved work:** workspaces and records used to make retries safe have no automatic expiry. Retry records can include earlier workspace content. They remain until the maintainer removes them, including in response to a verified removal request. There is currently no self-service delete control or fixed removal-response deadline. Disconnecting the plugin does not delete saved work.
+- **Connector authorization:** consent transactions and their temporary browser cookies expire after 10 minutes. Access tokens are valid for up to one hour; refresh grants expire 30 days after issuance. Expired database records cannot be used and are removed on a subsequent OAuth storage write, so expiry is not a guaranteed physical deletion time. Client registrations have no automatic expiry and remain until removed by the maintainer.
+- **Usage counts:** daily aggregate counters are retained indefinitely; no automatic deletion schedule is configured. Turning reporting off stops future opted-out measurements and does not remove earlier aggregate counts. The counters contain no account identifier with which to locate an individual's contribution.
+
+ChatGPT Sites manages its own sign-in, native connection records and operational data. Its providers' retention and backup timelines are governed by their policies; Forward Deployed does not set or promise a deletion period for those copies.
+
+You can keep work stateless, export it, clear local browser storage, turn usage reporting off, or disconnect the connector in your AI host. For privacy questions or removal requests, use the [repository](https://github.com/anotb/forward-deployed) to ask the maintainer for a private contact channel. Keep that initial request nonsensitive: do not include account identifiers, client material, credentials or other personal information in a public issue. Account verification and details of the records to remove should be shared only through the agreed private channel.