Simplify Knapper agent sessions

[slate-rehm]

Summary

• Replace agent and workspace handles with one stateless managed session.
• Publish a fixed startup tool surface and serialize all operations through one FIFO lane.
• Add cross-process activity ownership with stale detection and active-operation heartbeats.
• Add a live plugin settings fixture that verifies a real UI toggle and state readback.
• Remove legacy parallel workspace and lease code.
• Prepare version 0.7.0-beta.1.

Verification

• npm run check
• npm run typecheck
• npm test: 465 tests
• npm run smoke
• npm run acceptance: 20 checks
• npm run e2e with settings fixture: 92 checks
• npm run workspaces: 7 checks
• npm run fence: 16 checks
• npm run bg-input: 6 checks
• Codex plugin validation
• Claude plugin validation

Note

Replace multi-workspace/agent session model with a single managed session in Knapper

• Removes the workspace/agent handle model entirely: tools no longer accept or inject workspaceHandle/agentHandle; session lifecycle is now managed via obsidian_session_open, obsidian_session_release, and obsidian_session_reset.
• Introduces ActivityGuard in activity-guard.ts as a cross-process coordination primitive that enforces a single active session at a time, surfacing KNAPPER_BUSY with retryAfterMs to competing callers.
• Serializes all tool invocations through a single FIFO lane in ToolRegistry (was configurable concurrency); status-only tools bypass activity acquisition.
• Removes the obsidian_toolsets_update tool; the MCP tool surface is now fixed at initialization with a default set of toolsets (core, ui, telemetry, plugin-dev, editor).
• Bumps version to 0.7.0-beta.1 and updates all documentation, skills, agent guides, and test scripts to reflect session semantics.
• Risk: all existing integrations passing workspaceHandle or relying on dynamic toolset updates or multi-workspace isolation will break.

^{Macroscope summarized 9d6a978.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 32ba660d-b2dd-4a68-b00c-bb8639269390

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

_{Comment @coderabbitai help to get the list of available commands.}

[slate-rehm]

#AI-Automation

CI passed without failures on Node 20, 22, and 24, including the packaged-install gate.

Review-driven changes:

• Preserved unknown toolset diagnostics when all is selected.
• Removed the remaining legacy registry constructor arguments.
• Added remediation to new typed errors.
• Added an active-operation heartbeat and safe heartbeat error handling.
• Kept invocation finalization inside the FIFO lock and added regression coverage.
• Prevented ownership release while a session claim or operation remains active.
• Made doctor and status calls report ownership without acquiring it.
• Corrected release, default-profile, toolset-default, and platform-isolation guidance.
• Aligned marketplace descriptions and session setup instructions.
• Resolved every Macroscope review thread.