--- name: projects description: Use for Signalsurf Project discovery, Project Context, Channels, Threads, decisions, Activity, Routines, members, delegations, or Working File authority. --- # Signalsurf Projects Before every request using this Skill, follow [version check and automatic update](../signalsurf/references/freshness.md), then use the current guidance. Read the [shared operating contract](../signalsurf/references/operating-contract.md) unless already loaded; a focused Skill uses the same context, actor, confirmation and result rules. Follow the `signalsurf` operating contract. A Project is one durable Play, its in-product Surfer is the DRI, and each Thread is the coordination ledger for one piece of work. 1. Use `get_workspace_context` if Workspace context is absent or stale. 2. Resolve the Project with `project_query` actions `list`, `search`, or `get`, and load `get_project_context` when entering/resuming/switching work. A known `threadId` for existing durable work may be read before execution; routine operations do not search Threads to leave a log. 3. Execute the requested work and verify its resource result. Selecting a Project does not imply creating a Thread. Completing an operation does not imply posting an update. 4. Apply the shared operating contract's one writeback route: no Project writeback, one reply to same-objective work, or a material/distinct new Thread after objective search. Use `project_manage` action `post_channel_message` with `wakeSurfer: false` for that chosen result; for a new root omit the existing target, using null only if the live schema requires it. Read [Project results](references/results.md) when including an external request summary, publishing a long document or revising that document. A recent preview alone cannot establish absence. 5. For requested delegated/background execution, use the same posting action with `wakeSurfer: true`, an exact target and the bounded task, then inspect the actual delegation with `commander_query`. A request to put completed results in a Thread is post-only; it does not ask Surfer to perform the task again. 6. Use `project_file_authority` only for explicitly requested Project File grants, revision history or conflict work. Claim adds an independent write grant for that Project; release removes only that grant, and transfer is unsupported. Never add a grant automatically to make context selection succeed. Routine reads do not require a lease. Project messages retain the authenticated member and external-client provenance. A request summary is client-reported context; the result is the client's contribution. Do not speak as Surfer, impersonate an exact human message or paste the private host transcript. Inspect Project Operation access and effective approval policy before business work. Inbox, Meetings and Content selection is separate from Workspace entitlement, OAuth and actual account/resource access; always_allow still follows native confirmations. The context snapshot is bounded; inspect additional Files/Threads only when needed and refresh after relevant settings change. Read Project File Access with project_query action access when selection is missing or unclear. A new Project starts without selected Operations or Records collections. Change an explicitly requested Inbox/messages, Meetings/meeting, Content/content or Records collection selection with project_manage action access and its exact confirmation; Project manager authority is required, and Records selection also requires Workspace Admin. Do not claim MCP cannot change these settings or send the user to the UI when this capability is available. An attached List exposes its backing Records dependency for that List's members; it does not select the whole collection or unrelated Records. Removing a List preserves independently selected Records. Collection-wide edits require explicit collection access; List enrichment bindings, reads and exact member runs use their validated List scope. For requested one-time, recurring or autonomous Project work, read [schedule routing](../signalsurf/references/schedule-routing.md). It owns native Routine configuration, AI participation, scope, timing/event conditions, audience, persistence and execution proof; a posted Thread or saved active schedule is not a completed run. For a Routine's Related members, read project_query action members and configure audienceUserIds using current human Project member IDs. The schedule routing reference covers exact mutation fields, one-time tasks that already fired and the distinction between saved configuration and execution proof.